Datasets:

---

feedback-to-code

/

mozzarella

like 1

Follow

Feedback-2-Code 7

[ "timestamp(timedelta=0.0, similarity=0.8566144616384757)" ]

[ "This is a reasonable feature request.\r\nShould be easy to implement." ]

[ "Why are we allocating a copy of the list here, only to throw it away?", "Python allows only string or tuple-of-string here, not list of string, and so should we.", "Done", "Done", "I believe we don't check the element type, that the `generic1` part of `ParamType` is just for documentation generation purposes. This should be replaced by\r\n\r\n for (String s : subs.getContents(String.class, \"string\")) {\r\n ...\r\n }\r\n\r\nSame for `startsWith` below.", "Can we add cases for:\r\n- singleton tuple\r\n- empty tuple\r\n- tuple with element of wrong type", "Could be a good idea to document that.", "Done", "Indeed, I'll add that caveat to Param's javadoc now.\r\n\r\nNote that the var `subs` and the `@SuppressWarnings` are no longer needed, since the cast is done by `getContents`. `ConversionException` extends `EvalException` so this can just be `EvalException`. (Same below, of course.)", "Thanks." ]

[ "type: feature request", "good first issue" ]

[ "src/main/java/com/google/devtools/build/lib/syntax/StringModule.java" ]

[ "src/main/java/com/google/devtools/build/lib/syntax/StringModule.java" ]

[ "src/test/skylark/testdata/string_misc.sky" ]

[ "timestamp(timedelta=147727.0, similarity=0.8530386804142502)" ]

[ "cc @katre", "I think the only way to handle this will be to use two toolchain types: one for the bootstrapping, one for actual usage. If you write the implementation functions modularly, and use the same property names on the toolchain providers, you should be able to re-use a lot of code.\r\n\r\nI'm closing this issue because I don't think this is something Bazel can support at this time." ]

[]

[ "type: feature request", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/DependencyResolver.java" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/DependencyResolver.java" ]

[ "src/test/shell/bazel/toolchain_test.sh" ]

[ "timestamp(timedelta=0.0, similarity=0.8992251828312009)" ]

[]

[]

[]

[ "site/docs/tutorial/java.md" ]

[ "site/docs/tutorial/java.md" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.8610414419473065)" ]

[ "Thanks! If you feel like fixing it, here's the source for the doc file: https://github.com/bazelbuild/bazel/blob/master/site/docs/build-ref.html", "(Otherwise someone else will fix it later.)" ]

[]

[ "type: bug", "type: documentation (cleanup)" ]

[ "site/docs/build-ref.html" ]

[ "site/docs/build-ref.html" ]

[]

[ "timestamp(timedelta=77901.0, similarity=0.845440601490151)" ]

[ "Please provide the contents of jvm.out.", "> Please provide the contents of jvm.out.\r\n\r\nI ssh'd into CircleCI when this failure occurs and checked jvm.out and it an empty file after the crash. The failure looks like this:\r\n\r\n```\r\n[1] Segmentation fault\r\n[0] \r\n[0] Server terminated abruptly (error code: 14, error message: '', log file: '/home/circleci/.cache/bazel/_bazel_circleci/9ce5c2144ecf75d11717c0aa41e45a8d/server/jvm.out')\r\n[0] \r\n[0] bazel run //src:prodserver exited with code 37\r\n```\r\n\r\nIt doesn't always occur in the same place (there are a few steps in CI that I've seen it fail at) and on occasion it passes.\r\n\r\nIs there anywhere else besides jvm.out that might give a hint on the crash?", "If the file is empty then that's a sign that the gRPC client/server connection crashed. We've seen a similar case before where we didn't implement flow control correctly for file uploads to remote machines.\r\n\r\nDo you have remote caching or execution enabled? If yes, please provide the flags you are using.", "If these are the right flags, then it doesn't look like it:\r\n\r\n```\r\n[0] INFO: Options provided by the client:\r\n[0] Inherited 'common' options: --isatty=0 --terminal_columns=80\r\n\r\n[0] INFO: Reading rc options for 'run' from /home/circleci/ng/tools/bazel.rc:\r\n[0] Inherited 'build' options: --strategy=TypeScriptCompile=worker --strategy=AngularTemplateCompile=worker --symlink_prefix=dist/\r\n\r\n[0] INFO: Reading rc options for 'run' from /etc/bazel.bazelrc:\r\n[0] Inherited 'build' options: --noshow_progress --announce_rc --experimental_strict_action_env --experimental_repository_cache=/home/circleci/bazel_repository_cache --local_resources=3072,2.0,1.0\r\n```", "Hmm, that's odd - this seems to be a null build. Here's a more complete snippet from a passing run:\r\n```\r\n$ bazel build test/...\r\n...\r\nINFO: Analysed target //test/e2e:e2e (1 packages loaded).\r\nINFO: Found 1 target...\r\nTarget //test/e2e:e2e up-to-date:\r\n dist/bin/test/e2e/app.spec.d.ts\r\n...\r\n$ yarn e2e-prodserver && yarn e2e-devserver\r\n$ concurrently \"bazel run //src:prodserver\" \"while ! nc -z 127.0.0.1 5432; do sleep 1; done && protractor\" --kill-others --success first\r\n...\r\n[0] INFO: Analysed target //src:prodserver (1 packages loaded).\r\n[0] INFO: Found 1 target...\r\n[0] Target //src:prodserver up-to-date:\r\n[0] dist/bin/src/prodserver_bin.sh\r\n[0] dist/bin/src/prodserver\r\n[0] INFO: Elapsed time: 15.070s, Critical Path: 9.76s\r\n[0] INFO: 2 processes: 1 processwrapper-sandbox, 1 worker.\r\n[0] INFO: Running command line: dist/bin/src/prodserver src -p 5432\r\n...\r\n[0] history-server listening on port 5432; Ctrl+C to stop\r\n...\r\n[0] bazel run //src:prodserver exited with code null\r\n$ concurrently \"bazel run //src:devserver\" \"while ! nc -z 127.0.0.1 5432; do sleep 1; done && protractor\" --kill-others --success first\r\n...\r\n[0] INFO: Analysed target //src:devserver (0 packages loaded).\r\n[0] INFO: Found 1 target...\r\n[0] Target //src:devserver up-to-date:\r\n[0] dist/bin/src/devserver.MF\r\n[0] dist/bin/src/scripts_devserver.MF\r\n[0] dist/bin/src/devserver\r\n[0] INFO: Elapsed time: 4.219s, Critical Path: 0.07s\r\n[0] INFO: 0 processes.\r\n[0] INFO: Running command line: dist/bin/src/devserver\r\n[0] Server listening on http://ad2f21ad8e4a:5432/\r\n```", "The failing build fails to run //src:prodserver, but it's still just two actions.", "I think the reference to #3645 is a red herring.", "If I understand correctly, then you're running the containers with a 4 GB limit. --local_resources does not actually restrict Bazel's own memory usage, but maybe that's the intent?\r\n\r\nIIRC, Bazel is set to 4 GB max memory, which is unaffected by --local_resources. Maybe Bazel is trying to allocate too much memory and the container is killing it? Technically, this doesn't imply that Bazel's memory use has actually increased - we might be using a slightly different gc configuration or allocate memory more rapidly, and that might push it over the limit.\r\n\r\nIf that's correct, then you should try adding --host_jvm_args=-Xmx2G to the Bazel invocation, like this:\r\n```\r\nbazel --host_jvm_args=-Xmx2G build ...\r\n```\r\nor like this in the bazelrc:\r\n```\r\nstartup --host_jvm_args=-Xmx2G\r\n```", "Thanks for investigating, Ulf!\r\n\r\nYour simple explanation is right. The build passes three times in a row with the 2G heap limit for the bazel JVM. https://circleci.com/gh/alexeagle/workflows/angular-bazel-example/tree/test-bazel-0.14.1\r\n\r\nI also have a PR out to update Angular to use 0.14 again\r\nhttps://github.com/angular/angular/pull/24512\r\nwhere I've increased the size of the VM used to run one of the jobs.\r\n\r\nThis should fix it for us. I'm not sure what else we could do on this issue, other than improve the \"guard rails\" here so it's harder to get the wrong memory limit, or easier to debug. Feel free to close if you don't want to take further action." ]

[ "can assert_size itself be a test? then you don't need this extra rule, and dummy.sh can be hidden away.", "Do nothing, feel better?", "This is an interesting approach. What was the reasoning behind? assert_size is a nice Skylark rule, it only feels weird to be used like this in a BUILD file. Can you create and run an assert_size() target from the sh test?", "> can assert_size itself be a test?\r\n\r\nYes it can! Done.\r\n\r\n> Can you create and run an assert_size() target from the sh test?\r\n\r\nThat'd beat the purpose of analysis-time assertion and of not building the inputs. See the new comments in the docstring.", "Now even less! (Removed.)", "Oh, I see now! That's smart!" ]

[ "under investigation", "P1" ]

[ "src/BUILD" ]

[ "src/BUILD", "src/rule_size_test.bzl" ]

[]

[ "timestamp(timedelta=80842.0, similarity=0.8454423582900445)" ]

[ "I think this is all fixed now." ]

[ "Try not to introduce more `using std::type` aliases. When reader sees `mutex` in a function, it is not immediately clear if it is `std::mutex` or Bazel's own custom mutex implementation.\r\n\r\n`using MagicMap = std::unordered_map<std::string, MagicItem>` that saves significant amount of typing is strongly encouraged.", "Done. Thanks for the advice!" ]

[ "type: bug", "P1", "type: process", "category: misc > misc" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/blaze_util_platform.h", "src/main/cpp/blaze_util_posix.cc", "src/main/cpp/blaze_util_windows.cc" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/blaze_util_platform.h", "src/main/cpp/blaze_util_posix.cc", "src/main/cpp/blaze_util_windows.cc" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9050090693317487)" ]

[]

[]

[ "type: bug" ]

[ "src/main/java/com/google/devtools/build/lib/vfs/JavaIoFileSystem.java", "src/main/java/com/google/devtools/build/lib/windows/BUILD", "src/main/java/com/google/devtools/build/lib/windows/WindowsFileSystem.java", "src/main/java/com/google/devtools/build/lib/windows/jni/WindowsFileOperations.java" ]

[ "src/main/java/com/google/devtools/build/lib/vfs/JavaIoFileSystem.java", "src/main/java/com/google/devtools/build/lib/windows/BUILD", "src/main/java/com/google/devtools/build/lib/windows/WindowsFileSystem.java", "src/main/java/com/google/devtools/build/lib/windows/jni/WindowsFileOperations.java" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.931227190370861)" ]

[]

[]

[]

[ "src/tools/skylark/java/com/google/devtools/skylark/skylint/BadOperationChecker.java", "src/tools/skylark/javatests/com/google/devtools/skylark/skylint/BadOperationCheckerTest.java" ]

[ "src/tools/skylark/java/com/google/devtools/skylark/skylint/BadOperationChecker.java", "src/tools/skylark/javatests/com/google/devtools/skylark/skylint/BadOperationCheckerTest.java" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.9596030179739292)" ]

[ "Looking at this, support should be pretty simple it seems? Even almost working as-is?", "Looking at the commit, never mind 😄 " ]

[]

[ "type: feature request", "P2", "team-Android" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/AndroidNdkRepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/StlImpl.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/StlImpls.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/AndroidNdkRepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/StlImpl.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/StlImpls.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/AndroidNdkCrosstoolsR18.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/ApiLevelR18.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/NdkMajorRevisionR18.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/X86Crosstools.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9743079330526386)" ]

[ "@meisterT for triage since this is perf related. I do agree that \"every 10 seconds full gc\" is not awesome.", "I think we should delete this code. I have yet to see a benchmark showing that the application logic choosing to do a GC results in better performance than the JVM triggering the GC itself." ]

[ "I think it's better not to create the IdleServerTasks if we don't have to, so perhaps rename the option and check for nullness below. WDYT?", "Contrary to my earlier statement, I think it's safer to default to the old behavior. Then, we can release it with the toggle and flip the default value in a later release.", "You mean rename the command line option or something else?\r\n\r\nThe name `IdleServerTasks` implies a generic framework for executing tasks at idle not just GC. Of course, it's not currently used for anything else. So, if I make the refactor I think you're suggesting, maybe I should just rename it `IdleGCTaskRunner`.", "That makes sense.", "My suggestion is to keep the name IdleServerTasks and the generic framework.\r\nBut instead of disabling the GC part of it with --idle_gc=no, I would disable the whole IdleServerTasks with --idle_server_tasks=no. " ]

[ "type: feature request" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/startup_options.cc", "src/main/cpp/startup_options.h", "src/main/java/com/google/devtools/build/lib/runtime/BlazeRuntime.java", "src/main/java/com/google/devtools/build/lib/runtime/BlazeServerStartupOptions.java", "src/main/java/com/google/devtools/build/lib/server/GrpcServerImpl.java", "src/main/java/com/google/devtools/build/lib/server/IdleServerTasks.java", "src/main/java/com/google/devtools/build/lib/server/RPCServer.java" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/startup_options.cc", "src/main/cpp/startup_options.h", "src/main/java/com/google/devtools/build/lib/runtime/BlazeRuntime.java", "src/main/java/com/google/devtools/build/lib/runtime/BlazeServerStartupOptions.java", "src/main/java/com/google/devtools/build/lib/server/GrpcServerImpl.java", "src/main/java/com/google/devtools/build/lib/server/IdleServerTasks.java", "src/main/java/com/google/devtools/build/lib/server/RPCServer.java" ]

[]

[ "timestamp(timedelta=130064.0, similarity=0.8452684891346447)" ]

[ "To confirm, is the request to support using the embedded JDK as the `--javabase`?\r\n\r\nIf the goal is to have `--javabase` versioned with your respository, have you considered versioning a separate JDK?", "Not quite. What I'm really looking for is a way to completely eliminate uses of any JDKs found via `which javac` or `$JAVA_HOME` or any other form of looking around on the system. Experimenting with `--javabase=@embedded_jdk//:jdk --host_javabase=@embedded_jdk//:jdk` just now, builds still fail on a machine with a weird (broken?) JDK exposed via `which javac` because that JDK is missing jar files in places bazel expects them.\r\n\r\nI would also like a way to use the embedded JDK as the alternative so I have less stuff to version, but I wouldn't mind versioning a separate one via a bazel package too much.\r\n\r\nAs far as I can see, there are no user-accessible flags to change `@local_jdk`. src/main/cpp/blaze.cc always passes its own value for `--default_system_javabase`, and then everything else just uses that value. If there was some way to do that, I think it would be enough for my use case. I guess I could make my tools/bazel wrapper script download a fixed JDK, extract it, and set JAVA_HOME to point to it, but that seems like an awful lot of infrastructure to add to emulate something that used to just work. It's also a super trivial 1-line patch to make `StartupOptions::GetSystemJavabase()` return an empty string, and seems like it wouldn't be much harder to control that via a flag.", "> > To confirm, is the request to support using the embedded JDK as the --javabase?\r\n>\r\n> Not quite. What I'm really looking for is a way to completely eliminate uses of any JDKs found via which javac or $JAVA_HOME or any other form of looking around on the system.\r\n\r\nTo try to recap, there are three JDKs:\r\n1) the JDK used to run Bazel itself. This is the `startup --host_javabase` (i.e. the LHS host_javabase)\r\n2) the JDK used to run host tools that run during the build, e.g. javac. This is the `build --host_javabase` (i.e. the RHS host_javabase)\r\n3) the JDK that the build is targeting, that is used to run any `java_binary` and `java_tests` that a build produces. This is the `--javabase`.\r\n\r\n(1) and (2) now default to the embedded JDK (which is not the local JDK). They should be unaffected by `JAVA_HOME` and the presence of `javac` on the path.\r\n\r\n(3) is the one that depends on `JAVA_HOME` / `javac`. You can set it explicitly using `--javabase`. `--javabase=@embedded_jdk//:jdk` doesn't work because Bazel doesn't support using JDK 9 as a target `--javabase` yet.\r\n\r\nIf you obtain a separate JDK 8, create a `java_runtime` rule and set the `java_runtime.java_home` attribute to the root of that JDK 8, and then set `--javabase` to the label of the `java_runtime` rule, that should guarantee that Bazel's behaviour is unaffected by `JAVA_HOME` etc.\r\n\r\nDoes that match what you're seeing?\r\n\r\n> I would also like a way to use the embedded JDK as the alternative so I have less stuff to version\r\n\r\nThis makes sense, but it isn't really the intended use-case for the embedded JDK. There's some discussion about shipping a minimal JDK that contains only what's need for the host_javabase use-case, which probably isn't what's wanted fro a target `--javabase`. It would be less fragile to explicitly choose the JDK version to target instead of using whatever version Bazel happens to be currently using.", "@philwo @buchgr FYI\r\n\r\nThere's some discussion happening about making it possible to configure host JDKs using remote repositories or similar, and it would probably be possible to use the same approach for target JDKs. So if you want the ability to version the host and target JDKs and avoid implicitly depending on the local JDK that'd be another way to approach it.\r\n\r\nThis original feature request (of using the embedded JDK as a target JDK) is not something we're likely to pursue for the reason I mentioned above: we want to make the embedded JDK image as a small as possible to reduce the installation size.", "@cushon\nAre there tickets for these discussions (depending on an external versioned\njdk for host/target)?\nOn Thu, 26 Jul 2018 at 17:30 Liam Miller-Cushon <[email protected]>\nwrote:\n\n> @philwo <https://github.com/philwo> @buchgr <https://github.com/buchgr>\n> FYI\n>\n> There's some discussion happening about making it possible to configure\n> host JDKs using remote repositories or similar, and it would probably be\n> possible to use the same approach for target JDKs. So if you want the\n> ability to version the host and target JDKs and avoid implicitly depending\n> on the local JDK that'd be another way to approach it.\n>\n> This original feature request (of using the embedded JDK as a target JDK)\n> is not something we're likely to pursue for the reason I mentioned above:\n> we want to make the embedded JDK image as a small as possible to reduce the\n> installation size.\n>\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5594#issuecomment-408137102>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/ABUIFxQaUNJTbTzOh8lWcW80q_ySjISJks5uKeCsgaJpZM4VPYjF>\n> .\n>\n", "@ittaiz https://groups.google.com/d/topic/bazel-sig-jvm/noXlM1HOvVY/discussion", "Even if I set --javabase and --host_javabase, builds still don't work on a machine where `which javac` finds a broken JDK. Is there some other flag that also needs to be set to completely avoid the system-installed JDK from affecting builds?", "Specifically, with this WORKSPACE:\r\n```python\r\nnew_http_archive(\r\n name = \"zulu_jdk8\",\r\n build_file = \"zulu_jdk8.BUILD\",\r\n sha256 = \"65a1b180ad46c5a1cc2f9ffff29b85d682c0e937b7b964b557f2a3b9084de5b9\",\r\n strip_prefix = \"zulu8.30.0.1-jdk8.0.172-linux_x64\",\r\n url = \"https://cdn.azul.com/zulu/bin/zulu8.30.0.1-jdk8.0.172-linux_x64.tar.gz\",\r\n)\r\n```\r\nAnd this BUILD:\r\n```python\r\ngenrule(\r\n name = \"gen_main\",\r\n outs = [\"Main.java\"],\r\n cmd = \"echo \\\"class Main { public static void main(String[] args) {} }\\\" > $@\",\r\n)\r\n\r\njava_binary(\r\n name = \"main\",\r\n srcs = [\"Main.java\"],\r\n main_class = \"Main\",\r\n)\r\n```\r\n\r\nThis works (I'm doing this on a machine with a working JDK found via `which javac`):\r\n```console\r\n$ /tmp/bazel-0.15.2-linux-x86_64 build //...\r\nINFO: Build options have changed, discarding analysis cache.\r\nINFO: Analysed 2 targets (16 packages loaded).\r\nINFO: Found 2 targets...\r\nINFO: Elapsed time: 5.201s, Critical Path: 2.85s\r\nINFO: 2 processes: 1 linux-sandbox, 1 worker.\r\nINFO: Build completed successfully, 7 total actions\r\n```\r\nBut this fails:\r\n```console\r\n$ JAVA_HOME=/nonexistent /tmp/bazel-0.15.2-linux-x86_64 build //...\r\nWARNING: Running Bazel server needs to be killed, because the startup options are different.\r\nStarting local Bazel server and connecting to it...\r\n.....................\r\nERROR: /x1/home/brian/.cache/bazel/_bazel_brian/9a14994a41429a580e5cd8b6f1b379ce/external/bazel_tools/tools/jdk/BUILD:181:1: no such package '@local_jdk//': Expected directory at /nonexistent but it does not exist. and referenced by '@bazel_tools//tools/jdk:jdk'\r\nERROR: Analysis of target '//:main' failed; build aborted: Analysis failed\r\nINFO: Elapsed time: 6.264s\r\nINFO: 0 processes.\r\nFAILED: Build did NOT complete successfully (10 packages loaded)\r\n currently loading: @bazel_tools//src/tools/launcher ... (2 packages)\r\n```", "> Even if I set --javabase and --host_javabase, builds still don't work on a machine where which javac finds a broken JDK.\r\n\r\nAre you setting the `--host_javabase` build flag, or startup flag, or both?", "I was using the build flag because I didn't realize there was a startup one. What version of JDK is supported for the startup one? If I use a Java 8 one, it breaks because `src/main/cpp/blaze.cc` says `TODO(b/109998449): only assume JDK >= 9 for embedded JDKs`, and then adds a Java 9-only flag for passing to the JVM found via the `--host_javabase` startup flag. If I use a Java 9 one, then compilation via the `java_*` rules breaks because they don't support Java 9 JDKs yet. Neither the `--host_javabase` or `--javabase` build flags seem to make any difference whatsoever because `@local_jdk` fails to load either way.", "> If I use a Java 9 one, then compilation via the java_* rules breaks because they don't support Java 9 JDKs yet.\r\n\r\nThe default `startup --host_javabase` is the embedded JDK, which is a JDK 10 in Bazel at head. It was JDK 9 until recently, and JDK 9 should still work.\r\n\r\nJDK 9 isn't supported as a target `--javabase` yet.\r\n\r\nIf you use `bazel --host_javabase=<jdk 9> build ...` that should work. If it doesn't, what error are you seeing?", "Oh, I was testing the wrong bazel binary... `bazel --host_javabase=<jdk 9> build ...` works with 0.6.0, but setting `JAVA_HOME=/dev/null/nonexistent` still breaks it. Same error as before:\r\n```console\r\nERROR: /x1/home/brian/.cache/bazel/_bazel_brian/efaa4e6bb8c9b0585e183b73359e618d/external/bazel_tools/tools/jdk/BUILD:149:1: no such package '@local_jdk//': Expected directory at /dev/null/nonexistent but it does not exist. and referenced by '@bazel_tools//tools/jdk:bootclasspath'\r\n```", "@buchgr @meisterT FYI. I think this is exactly the bug I found when we discussed ~2 weeks ago, which `--{host_,}javabase` flag influences what part of the build: Namely, the `@local_jdk` repo, which is *not* set to the path given by `--javabase` (which would be the right behavior), but is always taken from the JAVA_HOME environment variable (or calculated from `which javac`) and cannot be influenced at all via flags.\r\n\r\nHere's the current set of flags and what they do:\r\n\r\n- `bazel --host_javabase=... build ...` = Defines what JDK Bazel itself runs under.\r\n- `bazel build --host_javabase=...` = Defines the JDK that Bazel uses to run Java tools that run on your machine as part of the build (e.g. JavaBuilder, SingleJar and friends).\r\n- `bazel build --javabase=...` = Defines what Bazel puts into the launcher script of java_binary and java_test targets as their Java runtime. So, if you run `bazel test --javabase=//my:jdk //:javatest`, then `//my:jdk` would be used to run the test.\r\n- `$JAVA_HOME` = what is made available as `@local_jdk` and used by JavaBuilder as the classpath to build against.\r\n\r\nAFAIU the proposed fix is to merge the last two and make `bazel build --javabase=$JAVA_HOME` the default, so that people who have a JDK installed can just go and build stuff, but when you explicitly set `bazel build --javabase=//my:jdk`, then that would also redirect `@local_jdk` to `//my:jdk` and your system JDK (if it exists) would be completely ignored.\r\n\r\n(@cushon Please correct me if I'm wrong!)\r\n", "@philwo that all sounds right to me.\r\n\r\nIf I'm following, the goal should be that `@local_jdk` is used only as the default value of `--javabase`, and that all other uses of it (e.g. in `@bazel_tools//tools/jdk:bootclasspath` should be migrated to use `current_java_runtime`). By default the host-config `current_java_runtime` will still be `@local_jdk`, but changing `--javabase` will be sufficient to avoid using `JAVA_HOME` anywhere.", "That sounds like it would work for my use case. Any ETA on implementing it? I'm currently stuck not being able to roll out a new bazel release with other helpful bugfixes and new features due to this.", "cc @buchgr \r\n\r\n> That sounds like it would work for my use case. Any ETA on implementing it?\r\n\r\nI'm re-opening this bug to track breaking dependencies on `JAVA_HOME` / `@local_jdk` except as the default value of `--javabase`. I'm hoping that will be ready 'soon', but Bazel team may have other ideas to mitigate this issue.", "This is fixed at head, and the fix will be included in the upcoming 0.17 release.", "Almost. https://github.com/bazelbuild/bazel/issues/6341", "In case someone else gets confused about how to refer to the java toolchain from genrules etc., please have a look here: https://stackoverflow.com/questions/53066974/how-can-i-use-the-jar-tool-with-bazel-v0-19", "copying jdk tools.jar introduced bugs. refer to, https://github.com/bazelbuild/intellij/issues/2164" ]

[]

[ "type: feature request" ]

[ "tools/jdk/BUILD", "tools/jdk/DumpPlatformClassPath.java" ]

[ "tools/jdk/BUILD", "tools/jdk/DumpPlatformClassPath.java" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.8469261201945434)" ]

[ "Thanks for the report. I don't plan to work on this until https://github.com/bazelbuild/bazel/issues/5187 is done. And I can see an universe in which we won't need so many env vars to tweak C++ toolchain. But nothing concrete yet. If this is urgent, let me know, and I will help the PR.", "+1 to fixing this. This blocks using non-standard toolchains with bazel.\r\nE.g. in our experiments with downloading a toolchain on-the-fly in Tensorflow, we can't currently use the LLD linker from the downloaded package, because -B flag is added by bazel and can't be overriden (so we end up picking LLD in /usr/bin/ if it's installed).\r\n\r\nMy intuition is that there should be a fix for https://github.com/bazelbuild/bazel/issues/760 that does not involve setting -B." ]

[]

[ "type: feature request", "P2", "team-Rules-CPP" ]

[ "tools/cpp/CROSSTOOL", "tools/cpp/unix_cc_configure.bzl" ]

[ "tools/cpp/CROSSTOOL", "tools/cpp/unix_cc_configure.bzl" ]

[ "src/test/java/com/google/devtools/build/lib/analysis/mock/MOCK_CROSSTOOL", "src/test/shell/bazel/testdata/bazel_toolchain_test_data/tools/arm_compiler/CROSSTOOL" ]

[ "timestamp(timedelta=1.0, similarity=0.9344559452588765)" ]

[ "Thanks for reporting it!" ]

[ "This link downloads a \"visualcppbuildtools_full.exe\" which seems pretty confusing to say the least. Maybe we should link the page where all the Visual C++ Build Tools are?", "I agree. Microsoft is notorious for orphaning download links.", "Done." ]

[ "type: documentation (cleanup)", "platform: windows", "bazel.build" ]

[ "site/docs/windows.md" ]

[ "site/docs/windows.md" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9462255836554451)" ]

[]

[]

[ "type: bug", "P1", "platform: windows" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[]

[ "timestamp(timedelta=96914.0, similarity=0.9177194223528519)" ]

[ "Workaround for the time being (because of https://github.com/bazelbuild/bazel/issues/5802):\r\n- add a new `java_library` to `src/tools/runfiles/java/com/google/devtools/build/runfiles/BUILD.tools` called \"runfiles-deprecated\" with a `deprecation` attribute\r\n- change the `alias` rule in `tools/runfiles/BUILD.tools` to point to \"runfiles-deprecated\"\r\n- add a new `java_library` to `tools/java/runfiles/BUILD.tools` that uses the sources from `//src/tools/runfiles/java/com/google/devtools/build/runfiles:java-srcs`.", "This [workaround](https://github.com/bazelbuild/bazel/issues/5803#issuecomment-411388225) is not good enough: it warns against the alias rule, not against the rule that depends on the alias, i.e. Bazel will warn against `@bazel_tools//tools/runfiles:java-runfiles`, not against the rule that depends on `@bazel_tools//tools/runfiles:java-runfiles` (as we would like to)." ]

[ "Clever!" ]

[ "type: process" ]

[ "src/tools/runfiles/java/com/google/devtools/build/runfiles/BUILD", "src/tools/runfiles/java/com/google/devtools/build/runfiles/BUILD.tools", "src/tools/runfiles/java/com/google/devtools/build/runfiles/Runfiles.java", "tools/BUILD", "tools/runfiles/BUILD.tools" ]

[ "src/tools/runfiles/java/com/google/devtools/build/runfiles/BUILD", "src/tools/runfiles/java/com/google/devtools/build/runfiles/BUILD.tools", "src/tools/runfiles/java/com/google/devtools/build/runfiles/Runfiles.java", "tools/BUILD", "tools/java/BUILD", "tools/java/runfiles/BUILD", "tools/java/runfiles/BUILD.tools", "tools/runfiles/BUILD.tools" ]

[ "src/test/py/bazel/testdata/runfiles_test/bar/BUILD.mock", "src/test/py/bazel/testdata/runfiles_test/foo/BUILD.mock" ]

[ "timestamp(timedelta=101070.0, similarity=0.878287860173537)" ]

[ "@filipesilva I have made some progress on this recently.\r\nNow with Bazel built from HEAD, symlink tree will be available on Windows when you pass `--experimental_enable_runfiles`. \r\nCan you try if it helps with https://github.com/bazelbuild/bazel/issues/5926?", "@filipesilva Just found a bug when dealing with long path, sending a fix now." ]

[ "`FILE_FLAG_BACKUP_SEMANTICS` is required when and only when opening a directory or junction. When opening a file symlink, you must not use `FILE_FLAG_BACKUP_SEMANTICS`. I suggest trying to open it with `FILE_ATTRIBUTE_NORMAL | FILE_FLAG_OPEN_REPARSE_POINT` in case this `CreateFileW` call fails.\r\n\r\nFurthermore, in order to minimize risk of filesystem races I suggest attempting to open the file without trying to call `GetFileAttributesW`. The benefit: if the path does not exist, `CreateFileW` fails and `GetLastError()` will return a corresponding error code, but if the path does exist, we have an open handle to it and no other process can delete or modify the file until we release this handle. (The OS will delay actual deletion until all handles are closed.)\r\n\r\nAnd in case you do need the attributes after opening the file, you can use `GetFileInformationByHandle`.", "Good suggestions! Thanks!", "In my experience, CreateFileW(..., FILE_ATTRIBUTE_NORMAL | ...) fails for directories (and junctions, which are also directories), therefore this failure may just mean that \"path\" refers to a directory.", "Then I think we can call `GetFileAttributesW` first to check if the `path` refers to a reparse point or not. If it does, we can just use `FILE_FLAG_OPEN_REPARSE_POINT`, otherwise, we skip and return the original path. WDYT?", "And we can tell from the attribute if the `path` is a directory or not, if it is, we set `FILE_FLAG_BACKUP_SEMANTICS`", "Yes, I think calling GetFileAttributeW then CreateFileW (with FILE_ATTRIBUTE_NORMAL or FILE_FLAG_BACKUP_SEMANTICS, depending on the attributes) is as good as calling CreateFileW first with FILE_FLAG_BACKUP_SEMANTICS then with FILE_ATTRIBUTE_NORMAL -- both approaches achieve the same goal and both are susceptible to file system races (i.e. when the file is modified between the two function calls).", "OK, it's done. But I think we don't need `FILE_ATTRIBUTE_NORMAL` for file symlink because in the document it says `This attribute is valid only if used alone.`\r\nhttps://docs.microsoft.com/en-us/windows/desktop/api/fileapi/nf-fileapi-createfilea" ]

[ "type: feature request", "P2", "platform: windows", "area-Windows", "team-OSS" ]

[ "src/main/cpp/util/file_platform.h", "src/main/cpp/util/file_windows.cc" ]

[ "src/main/cpp/util/file_platform.h", "src/main/cpp/util/file_windows.cc" ]

[]

[ "timestamp(timedelta=2227.0, similarity=0.8835729751180569)" ]

[ "@filipesilva I have made some progress on this recently.\r\nNow with Bazel built from HEAD, symlink tree will be available on Windows when you pass `--experimental_enable_runfiles`. \r\nCan you try if it helps with https://github.com/bazelbuild/bazel/issues/5926?", "@filipesilva Just found a bug when dealing with long path, sending a fix now." ]

[ "Have you thought of using a `typedef`? I think that'd be more type-safe and more readable than a macro.", "Have you thought of using an unnamed namespace? That restricts visibility and the binary will not export these symbols.\r\n\r\n(See https://google.github.io/styleguide/cppguide.html#Unnamed_Namespaces_and_Static_Variables)", "What's the reason for using `explicit`?", "The error message is confusing, \"INPUT RUNFILES\" can be understood as \"a list of input runfiles\". What's the intended meaning? Could you clarify the message?", "I think the argument type does not need to be a const reference. Can you think of a more efficient approach?", "What does the rest of the loop do when `use_metadata` is false?", "What's the reason to store an entry when `target` is empty?", "Is this call necessary? Doesn't `ifstream`'s destructor close the file?", "String comparison should ignore casing.", "This call fails if the directory is not empty. Is that expected?", "Why can we ignore it?", "Have you considered using a foreach-loop?", "`CreateFileW` returns `INVALID_HANDLE_VALUE` on error. Please compare `h` to that instead of to `0`.", "type-safety: use `FALSE` `FALSE` (BOOL) instead of `false` (bool).", "Do you need to do this before every test, or would it suffice to export these envvars just once?", "Why is there one more symlink on Windows?", "You are calling `readlink $i` multiple times. Could you cache its result?", "why is there one more symlink on Windows?", "same here about `readlink $i`", "Good idea. Replaced it with typedef and moved it into the class definition.", "Good to know that, thanks!", "Done", "I copied from the Linux version build-runfiles.cc, but I think it's unnecessary, so removed.", "If `use_metadata` is false, it means every line in the manifest is a runfile entry, then the loop will read each of them and add them in the manifest map. If `use_metadata` is true, it will skip all lines with even line number, because they are metadata lines.\r\nSee https://github.com/bazelbuild/bazel/blob/master/src/main/tools/build-runfiles.cc#L144", "Bazel sometimes need to create empty files under the runfiles tree. For example, for python binary, `__init__.py` is needed under every directory. Storing an entry with an empty `target` indicates we need to create such a file when creating the runfiles tree.", "Indeed, removed.", "`AsAbsoluteWindowsPath` will convert the path to a normalized path, than means both of them are already in lower case.", "Here we are deleting a symlink, so it won't fail even the target directory is not empty.", "Because if the directory is not empty, it means it contains some symlinks already pointing to the correct targets (we just called `ScanTreeAndPrune`). Then this directory shouldn't be removed in the first place. It's better to check if the directory is empty before deleting it, but there's not simple kernel API for that.", "Good idea! Done.", "Done.", "Done", "Indeed, it's confusing. Updated.", "Right, once is enough.", "Because for shell binary, we build both `bin` and `bin.exe`, but on Linux we only build `bin`", "Yes, done.", "Answered in above comment.", "Done.", "Thanks. Please add this as a code comment.", "Thanks! Could you add this explanation to the code?", "Thanks! The convention in manpages is to put mandatory arguments in angle brackets, e.g. `<manifest_file>`. I think doing that here would increase readability.", "The name `use_metadata` confuses me, it suggests that the function should use (or ignore) the metadata. WDYT? If you agree, can you think of a better name?", "Could you add this as a comment?", "Could you add this to the code as a comment?", "Thanks, please add it as a comment here too.", "Indeed, `ignore_metadata` is better.", "Just realized `use_metadata` is also the flag name and used some other places in Java code. Let me instead add a comment here to clarify.", "Done.", "Done", "Done!", "Done", "Done" ]

[ "type: feature request", "P2", "platform: windows", "area-Windows", "team-OSS" ]

[ "src/main/cpp/util/BUILD", "src/main/java/com/google/devtools/build/lib/analysis/config/BuildConfiguration.java", "src/main/tools/BUILD", "src/main/tools/build-runfiles-windows.cc" ]

[ "src/main/cpp/util/BUILD", "src/main/java/com/google/devtools/build/lib/analysis/config/BuildConfiguration.java", "src/main/tools/BUILD", "src/main/tools/build-runfiles-windows.cc" ]

[ "src/test/py/bazel/runfiles_test.py", "src/test/shell/integration/BUILD", "src/test/shell/integration/runfiles_test.sh" ]

[ "timestamp(timedelta=2799.0, similarity=0.9213884834542884)" ]

[ "Revamped in https://github.com/bazelbuild/bazel/commit/92b14d3f5ada6c857ba9a3cefbfcf7affaf16450" ]

[ "how about \"This tutorial does not require experience with Bazel or Android app development.\"", "hmm, this seems to contradict the Android Studio being in Prerequisites?", "\"Initialize the project's workspace\"?", "this reassurance feels a bit odd, since presumably people are here because they want to work with android code", "\"and has a WORKSPACE file ...\"", "\"the app\" -> \"your project\"?", "hmm how about \"The `WORKSPACE` file contains the references to external dependencies that are required to build the project (it can also be empty if your project is self-contains).\"", "\"Integrate with the Android SDK\"? (though \"Integrating with the Android SDK\" sounds better, but I think you're trying to keep the parallelism with verbs)", "\"This will use the Android SDK at the path specified by the ANDROID_HOME environment variable ...\"", "\", for example:\"", "it might be good to say why one might want to explicitly specify these attributes on the rule", "i'm pretty sure that the package name would more accurately be \"the path to the BUILD file relative to the WORKSPACE\"", "\"Note that Bazel's package hierarchy is conceptually different from the Java package hierarchy of your Android App ...\"?", "\"For the simple Android app in this tutorial, we'll make all the source files in `src/main/java` a single Bazel package\"", "\"instructions\" -> \"declarations\"?", "consider putting the app's manifest one level up in `main`", "\"you'll first use the android_library rule to tell Bazel to build an Android library module\" (i.e., maintaining the distinction between declaration and instruction)", "\"Locate\" sounds better", "\"stores\" sounds too fancy. \"puts\"?", "I recommend omitting the `$` prompt indicator (for all platforms, for sake of consistency).\r\nAFAIK it's not customary on Windows and may be confusing and considered part of the command.", "In the Command Prompt, `setx` affects subsequent sessions but not the current one. `set` affects the current session only.\r\nTheir semantics are also different:\r\n- `setx` understands quoting, so `setx FOO \"some value\"` is appropriate and stores the value without quotes\r\n- `set` does not understand quoting, so `set FOO=some value` is the right approach; `set FOO=\"some value\"` would include the quotes in the value\r\n\r\nIn Powershell, the command you specified affects subsequent sessions but not the current one.\r\nTo set an envvar for the current session, you have to run `$env:FOO=\"some value\"`.", "I recommend adding an example for Windows too. The key thing to call out is that `path` must use mixed-style ( see `cygpath --mixed`) paths, i.e. Windows path with forward slashes, e.g. `c:/path/to/sdk`.", "`$WORKSPACE` does not make sense on Windows. I recommend some other notation, e.g. `<workspace>/bazel-out`.", "done.", "removed the line. I'll add it back when we have other tutorials on Android Studio and testing. ", "done.", "done.", "done.", "done.", "I broke it the current paragraph up into two lines so it reads crispier.", "done.", "done.", "done.", "done.", "done.", "done.", "done.", "done.", "done. https://github.com/bazelbuild/examples/pull/79", "done.", "done.", "done.", "done, removed all `$` prompts.", "thank you for the clear explanation! I split the table to two: one to set it for the current session, and another to set it persistently.", "done. added a Windows example.", "I removed the prefix, and added a note about where these directories can be found." ]

[ "type: feature request", "P2", "team-Android" ]

[ "site/docs/tutorial/android-app.md" ]

[ "site/docs/tutorial/android-app.md" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8801452286846743)" ]

[ "@iirina,\nhave we removed the blockers from this?\nWe still haven't been able to move rules_scala to the new one AFAIR and\nunfortunately due to the strict-deps false negative we're also stuck with a\nrather old rules_scala version\n\nOn Thu, Aug 9, 2018 at 1:32 AM c-parsons <[email protected]> wrote:\n\n> This is a tracking issue for offering a migration solution for\n> --incompatible_disallow_legacy_javainfo\n>\n> —\n> You are receiving this because you are subscribed to this thread.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5821>, or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/ABUIF8onFHTeQvFHV5_D19wEkgVhoAEpks5uO2bmgaJpZM4V0xLv>\n> .\n>\n", "Any context on the blockers here? (Is there another github issue?)", "I got this error message:\r\n> create_provider is deprecated and cannot be used when --incompatible_disallow_legacy_javainfo is set. Please migrate to the JavaInfo constructor.\r\n\r\nDocumentation is not great (https://docs.bazel.build/versions/master/skylark/lib/java_common.html#create_provider doesn't even say it's deprecated) and it's not clear how to migrate the existing code.\r\n\r\n(cc @lberki)", "@iirina Can you help `rules_scala` and `rules_kotlin` migrate?", "@laurentlb yes. @ittaiz What can I do to help `rules_scala`?", "I use the deprecated create_provider API as follows:\r\n\r\n```\r\nnew_providers.append(\r\n java_common.create_provider(\r\n use_ijar = False,\r\n compile_time_jars = provider.compile_jars,\r\n runtime_jars = provider.runtime_output_jars,\r\n transitive_compile_time_jars = compile_time_jars,\r\n transitive_runtime_jars = runtime_jars,\r\n ),\r\n )\r\n```\r\n\r\nin order to remove specific jars from the compile and runtime classpath. I do not believe I can implement the same functionality with the JavaInfo constructor. If it's possible, please tell me how so that I can port my code before the deprecated create_provider is removed from Bazel. Thanks.\r\n", "Can you say a bit more why you need to remove specific jars? Maybe it will\nhelp think what you can do\nOn Fri, 25 Jan 2019 at 21:14 Robert Brown <[email protected]> wrote:\n\n> I use the deprecated create_provider API as follows:\n>\n> new_providers.append(\n> java_common.create_provider(\n> use_ijar = False,\n> compile_time_jars = provider.compile_jars,\n> runtime_jars = provider.runtime_output_jars,\n> transitive_compile_time_jars = compile_time_jars,\n> transitive_runtime_jars = runtime_jars,\n> ),\n> )\n>\n> in order to remove specific jars from the compile and runtime classpath. I\n> do not believe I can implement the same functionality with the JavaInfo\n> constructor. If it's possible, please tell me how so that I can port my\n> code before the deprecated create_provider is removed from Bazel. Thanks.\n>\n> —\n> You are receiving this because you were mentioned.\n>\n>\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5821#issuecomment-457690483>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/ABUIFwJRigGu2m6aO1-PQEevHBWv0nNzks5vG1eRgaJpZM4V0xLv>\n> .\n>\n", "Java developers expect to be able to override the library dependencies of the libraries their code depends upon. For instance, suppose my code depends on two libraries, lib1 and lib2. The lib1 target has a dependency on guava22 and the lib2 target depends on guava25. I want to use both in my application.\r\n\r\nMy Starlark code implements a rule called java_exclude_library that removes jars from the compilation and run-time Java classpaths. You use it like this:\r\n```\r\njava_exclude_library(\r\n name = \"lib1_without_guava\",\r\n excludes = [\r\n \"//third_party/com.google.guava/guava22\",\r\n ],\r\n deps = [\r\n \":lib1\",\r\n ],\r\n)\r\n```\r\nto create lib1_without_guava from java_library lib1. The lib1_without_guava target is identical to lib1, except that the jar for guava22 does not appear on its compilation and run-time classpath. My application depends on targets lib1_without_guava and lib2. All the code in my application ends up using guava25 at run time.\r\n", "I added a comment to the issue explaining why I need a JavaInfo constructor\nthat allows control over classpath jars.\n\nOn Sat, Jan 26, 2019 at 3:59 PM Ittai Zeidman <[email protected]>\nwrote:\n\n> Can you say a bit more why you need to remove specific jars? Maybe it will\n> help think what you can do\n> On Fri, 25 Jan 2019 at 21:14 Robert Brown <[email protected]>\n> wrote:\n>\n> > I use the deprecated create_provider API as follows:\n> >\n> > new_providers.append(\n> > java_common.create_provider(\n> > use_ijar = False,\n> > compile_time_jars = provider.compile_jars,\n> > runtime_jars = provider.runtime_output_jars,\n> > transitive_compile_time_jars = compile_time_jars,\n> > transitive_runtime_jars = runtime_jars,\n> > ),\n> > )\n> >\n> > in order to remove specific jars from the compile and runtime classpath.\n> I\n> > do not believe I can implement the same functionality with the JavaInfo\n> > constructor. If it's possible, please tell me how so that I can port my\n> > code before the deprecated create_provider is removed from Bazel. Thanks.\n> >\n> > —\n> > You are receiving this because you were mentioned.\n> >\n> >\n> > Reply to this email directly, view it on GitHub\n> > <https://github.com/bazelbuild/bazel/issues/5821#issuecomment-457690483\n> >,\n> > or mute the thread\n> > <\n> https://github.com/notifications/unsubscribe-auth/ABUIFwJRigGu2m6aO1-PQEevHBWv0nNzks5vG1eRgaJpZM4V0xLv\n> >\n> > .\n> >\n>\n> —\n> You are receiving this because you commented.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5821#issuecomment-457865518>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/AABqsUwdYr11YsbJKpNQSbwh24N7HKIaks5vHMGrgaJpZM4V0xLv>\n> .\n>\n", "👍🏽\nI think maybe the discussion on the specific (and interesting) use case\nshould continue on a separate issue but I’ll leave that decision to @iirina\nand @lberki\nOn Mon, 28 Jan 2019 at 18:16 Robert Brown <[email protected]> wrote:\n\n> I added a comment to the issue explaining why I need a JavaInfo constructor\n> that allows control over classpath jars.\n>\n> On Sat, Jan 26, 2019 at 3:59 PM Ittai Zeidman <[email protected]>\n> wrote:\n>\n> > Can you say a bit more why you need to remove specific jars? Maybe it\n> will\n> > help think what you can do\n> > On Fri, 25 Jan 2019 at 21:14 Robert Brown <[email protected]>\n> > wrote:\n> >\n> > > I use the deprecated create_provider API as follows:\n> > >\n> > > new_providers.append(\n> > > java_common.create_provider(\n> > > use_ijar = False,\n> > > compile_time_jars = provider.compile_jars,\n> > > runtime_jars = provider.runtime_output_jars,\n> > > transitive_compile_time_jars = compile_time_jars,\n> > > transitive_runtime_jars = runtime_jars,\n> > > ),\n> > > )\n> > >\n> > > in order to remove specific jars from the compile and runtime\n> classpath.\n> > I\n> > > do not believe I can implement the same functionality with the JavaInfo\n> > > constructor. If it's possible, please tell me how so that I can port my\n> > > code before the deprecated create_provider is removed from Bazel.\n> Thanks.\n> > >\n> > > —\n> > > You are receiving this because you were mentioned.\n> > >\n> > >\n> > > Reply to this email directly, view it on GitHub\n> > > <\n> https://github.com/bazelbuild/bazel/issues/5821#issuecomment-457690483\n> > >,\n> > > or mute the thread\n> > > <\n> >\n> https://github.com/notifications/unsubscribe-auth/ABUIFwJRigGu2m6aO1-PQEevHBWv0nNzks5vG1eRgaJpZM4V0xLv\n> > >\n> > > .\n> > >\n> >\n> > —\n> > You are receiving this because you commented.\n> > Reply to this email directly, view it on GitHub\n> > <https://github.com/bazelbuild/bazel/issues/5821#issuecomment-457865518\n> >,\n> > or mute the thread\n> > <\n> https://github.com/notifications/unsubscribe-auth/AABqsUwdYr11YsbJKpNQSbwh24N7HKIaks5vHMGrgaJpZM4V0xLv\n> >\n> > .\n> >\n>\n> —\n> You are receiving this because you were mentioned.\n>\n>\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5821#issuecomment-458195838>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/ABUIF0vDWtHoEvQKhu3NUp5ya3uuHjgjks5vHyJggaJpZM4V0xLv>\n> .\n>\n", "Here is a link to the ts_java_exclude_library Starlark rule that uses java_common.create_provider.\r\nI do not believe this rule can be implemented once the create_provider API is removed.\r\n\r\nhttps://github.com/twosigma/bazel-rules/blob/master/ts_java_exclude.bzl\r\n", "@brown Thanks for letting me know. Can you explain briefly what the rule is trying to accomplish, especially the method `_create_exclude_providers`? I'm trying to understand your use case and come up with a solution.\r\n\r\n", "You can control what ends-up on the runtime classpath using [JavaInfo(neverlink = ...)](https://docs.bazel.build/versions/master/skylark/lib/JavaInfo.html#JavaInfo). I'm not sure why you need to control the compilation classpath as well.", "I don't think the neverlink parameter to JavaInfo helps.\r\n\r\nThe goal is to create a Java library from other Java libraries, but with some transitive dependencies removed from the run-time class path.\r\n\r\nLet say one group develops Java library1, which has many transitive dependencies including a java_import of Guava version 19. \r\n\r\nA second group develops Java library2, which has many transitive dependencies including a java_import of Guava version 20.\r\n\r\nI want to create a Java binary that depends on both library1 and library2, but if my Java code depends on both libraries, then both versions of Guava will be on my application's class path, which can cause problems.\r\n\r\nI can use the ts_java_exclude_library rule to make a new library that combines both library1 and library2, but excludes Guava 19:\r\n\r\n ts_java_exclude_library(\r\n name = \"libraries_without_guava19\",\r\n excludes = [ \"//ext/public/google/guava/19/0:jars\" ],\r\n deps = [\r\n \":library1\",\r\n \":library2\",\r\n ],\r\n )\r\n\r\nAbove, the excluded target is the java_import of Guava 19 jar files that library1 transitively depends on.\r\n\r\nMy Java binary depends on libraries_without_guava19. It includes library1 and its dependencies, which have been compiled with Guava 19. It also includes library2 and its dependencies, which have been compiled with Guava 20. Both library1 and library2 use Guava 20 when my binary runs, because Guava 19 has been excluded from the class path.\r\n", "The _create_exclude_providers function iterates through a list of Java Providers. For each Provider it creates a new one that's identical to the original, except when specific jar files occur in the original's run-time or compile-time class paths. In that case the new Provider lacks the jars in its class paths.\r\n\r\nThe code may not need to remove jars from the compile-time class path.\r\n", "This wasn't flipped in time for 0.25. Changing label to 0.26. ", "> The goal is to create a Java library from other Java libraries, but with some transitive dependencies removed from the run-time class path.\r\n\r\nThis is what `neverlink` is supposed to do, and you can use it on any kind of java rule targets. Assuming the guava targets defined via java_import(name = 'guava19',...) and java_import(name = 'guava20',...), you can set `neverlink = true` on guava19 and it won't end up on the runtime classpath. Can you think of disadvantages of always setting a java_import/java_library as neverlink in your project? I'm more inclined towards a solution with higher abstraction rather than cherry picking classpath jars.", "@ittaiz What's the status on the scala rules? Is there anything I can help with to migrate to the JavaInfo constructor?", "I too would be happier with a API that does not involve directly modifying Java class paths, but I currently don't have any good ideas for implementing an exclude feature without the ability to manipulate class path lists.\r\n\r\nTwo Sigma has a large monorepo containing thousands of external Java libraries, most of which have been imported from Maven. The Maven dependencies between packages are programmatically converted into dependencies between Bazel java_import libraries. With these dependencies a Two Sigma library can depend on an external Java package and all of that package's run time needs appear on an application's Java class path. If some external libraries have neverlink set to True, then libraries and/or binaries that depend on them will have to use runtime_deps to add back required run time dependencies. Managing runtime_deps lists with hundreds of entries would be a nightmare.\r\n\r\nLet's say I took your suggestion and set neverlink to True for the java_import of library guava19. That could be used to remove guava19 from my application's or library's Java class path. But what about all the other binaries and libraries in my monorepo that prefer to use guava19? I have to track down those Bazel targets and add guava19 to the runtime_deps of binaries and/or libraries.\r\n\r\nThe deprecated JavaInfo constructor is the only Bazel API that allows a rule to remove a dependency from the Java class path once it has been added by a transitive dependency. In the Two Sigma monorepo there are hundreds of instances where a jar is excluded from a Java class path. Google doesn't encounter this problem primarily because of its one version policy for external library dependencies. In a company with a monorepo containing several versions of many external libraries, the ability to depend on an internal library but selectively exclude some of its transitive external dependencies is valuable.\r\n", "What is the migration status for this issue?", "@dslomov am I just 5 minutes late on every issue? :)", "@iirina @ittaiz do you have any context here? I'm unsure of what remains to be done for rules scala", "Friendly ping @ittaiz :)", "@brown The flag needs to be flipped in Bazel 1.0. I put together some snippets to help you migrate your use case and still be able to exclude some jars. Please let me know how I can help further.\r\n\r\n\r\n```\r\n# after computing runtime_jars and compile_time_jars\r\n\r\ntransitive_compile_providers = [JavaInfo(\r\n output_jar = compile_time_jar,\r\n compile_jar = compile_time_jar,\r\n neverlink = True,\r\n)\r\nfor compile_time_jar in compile_time_jars]\r\n\r\ntransitive_runtime_providers = [JavaInfo(\r\n output_jar = runtime_jar,\r\n compile_jar = runtime_jar,\r\n)\r\nfor runtime_jar in runtime_jars]\r\n\r\ncompile_providers = [JavaInfo(\r\n output_jar = compile_jar,\r\n compile_jar = compile_jar,\r\n neverlink = True,\r\n)\r\nfor compile_jar in provider.compile_jars]\r\n\r\nruntime_providers = [JavaInfo(\r\n output_jar = runtime_jar,\r\n compile_jar = runtime_jar,\r\n deps = transitive_runtime_providers,\r\n)\r\nfor runtime_jar in provider.runtime_output_jars]\r\n\r\nnew_providers.append(java_common.merge(compile_providers + runtime_providers)\r\n```", "Is there migration tooling available/possible? How bad is the breakage?", "@dslomov No, the migration is different from use case to use case. \r\n\r\nI migrated the internal projects and also `rules_scala` in https://github.com/bazelbuild/rules_scala/pull/781. There is an open PR for `rules_kotlin` in https://github.com/bazelbuild/rules_kotlin/pull/197, and one internal open change for migrating intellij. \r\nOther than that, `rules_jvm_external` have to upgrade their `rules_scala` and `rules_kotlin` versions (cc @jin). No other breakages on our CI, but I don't know if projects that are not on BazelCI have migrated.", "Thanks very much for the snippet of code. It seems to work well.\nI will patch the library rule. Thanks again!\n\n\nOn Mon, Aug 5, 2019 at 10:40 AM Irina Iancu <[email protected]>\nwrote:\n\n> @brown <https://github.com/brown> The flag needs to be flipped in Bazel\n> 1.0. I put together some snippets to help you migrate your use case and\n> still be able to exclude some jars. Please let me know how I can help\n> further.\n>\n> # after computing runtime_jars and compile_time_jars\n>\n> transitive_compile_providers = [JavaInfo(\n> output_jar = compile_time_jar,\n> compile_jar = compile_time_jar,\n> neverlink = True,\n> )\n> for compile_time_jar in compile_time_jars]\n>\n> transitive_runtime_providers = [JavaInfo(\n> output_jar = runtime_jar,\n> compile_jar = runtime_jar,\n> )\n> for runtime_jar in runtime_jars]\n>\n> compile_providers = [JavaInfo(\n> output_jar = compile_jar,\n> compile_jar = compile_jar,\n> neverlink = True,\n> )\n> for compile_jar in provider.compile_jars]\n>\n> runtime_providers = [JavaInfo(\n> output_jar = runtime_jar,\n> compile_jar = runtime_jar,\n> deps = transitive_runtime_providers,\n> )\n> for runtime_jar in provider.runtime_output_jars]\n>\n> new_providers.append(java_common.merge(compile_providers + runtime_providers)\n>\n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/5821?email_source=notifications&email_token=AAAGVMO5U7QV5EVP3IR3V5TQDA3VJA5CNFSM4FOTCLX2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD3SA3BY#issuecomment-518262151>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/AAAGVMOUKW3FOVV4ZLVKETTQDA3VJANCNFSM4FOTCLXQ>\n> .\n>\n", "@brown That's great to hear, thanks for trying it out!" ]

[]

[ "P2", "team-Rules-Java", "incompatible-change" ]

[ "src/main/java/com/google/devtools/build/lib/packages/StarlarkSemanticsOptions.java", "src/main/java/com/google/devtools/build/lib/syntax/StarlarkSemantics.java" ]

[ "src/main/java/com/google/devtools/build/lib/packages/StarlarkSemanticsOptions.java", "src/main/java/com/google/devtools/build/lib/syntax/StarlarkSemantics.java" ]

[ "src/test/java/com/google/devtools/build/lib/rules/java/JavaInfoSkylarkApiTest.java", "src/test/shell/bazel/bazel_java_test.sh" ]

[ "timestamp(timedelta=0.0, similarity=0.909201512789194)" ]

[]

[]

[ "team-Rules-Java", "untriaged" ]

[ "tools/build_defs/repo/jvm.bzl" ]

[ "tools/build_defs/repo/jvm.bzl" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.8811939021390516)" ]

[ "Suggestion: enforce that Identifiers are either\r\n * `lower_snake_case`,\r\n * `UPPER_SNAKE_CASE`, or\r\n * `UpperCamelCaseWithSuffixInfo`" ]

[ "Someone told me the message could be improved:\r\n\r\n> I think this is confusing, made me think of \"information about a suffix\", not of the literal suffix \"Info\".\r\n> How about \"UpperCamelCase and end in \\\"Info\\\" (for providers; example: FooBarInfo)\" ?" ]

[ "type: bug", "P2" ]

[ "site/docs/skylark/skylint.md", "src/tools/skylark/java/com/google/devtools/skylark/skylint/NamingConventionsChecker.java", "src/tools/skylark/javatests/com/google/devtools/skylark/skylint/NamingConventionsCheckerTest.java" ]

[ "site/docs/skylark/skylint.md", "src/tools/skylark/java/com/google/devtools/skylark/skylint/NamingConventionsChecker.java", "src/tools/skylark/javatests/com/google/devtools/skylark/skylint/NamingConventionsCheckerTest.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8711573184192934)" ]

[ "We love contributions @Profpatsch :-)", "> We love contributions @Profpatsch :-)\r\n\r\nI will open a pull request, do you have any preference on what the `perl` interpreter should replaced with? I think we can assume command line arguments will only use ASCII characters, is that correct?", "I don't but @aehlig might", "> perl -ne 'print uc'\n> \n> [...] As far as I understand this uppercases letters. It also introduces a hard dependency on perl to do bash completion. \n\nOh, wow. We're using perl, just to convert a string to upper case letters; and not even\nany string, as far as I can see this string is just used for commands and options---and those\nare, for sure, 7-bit clean prinable ASCII.\n\nIn the spirit of not introducing additional dependencies, I vote for the `tr a-z A-Z`\nsolution, as the script currently depends on `tr(1)` anyway.\n\n-- \nKlaus Aehlig\nGoogle Germany GmbH, Erika-Mann-Str. 33, 80636 Muenchen\nRegistergericht und -nummer: Hamburg, HRB 86891\nSitz der Gesellschaft: Hamburg\nGeschaeftsfuehrer: Paul Terence Manicle, Halimah DeLaine Prado\n" ]

[]

[ "untriaged", "team-Bazel" ]

[ "scripts/bazel-complete-template.bash" ]

[ "scripts/bazel-complete-template.bash" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9003379059109604)" ]

[ "I'm about to submit a pull request which fixes this issue.", "Here's the pull request: https://github.com/bazelbuild/bazel/pull/5953" ]

[ "I think the changes in this file are no longer necessary and can be reverted?", "I think we don't want to clear the pipeline but actually fail if there's anything in there.\r\nIt would be a bug in our code if the pipeline were not empty.", "Good point, I'll revert this back since the exception handler should cover the race condition universally.", "Okay, I've updated the two `acquire*Channel()` methods to check if the pipeline is not empty and fail the promise if so. The pipeline could be in a strange state if there are any handlers on a newly acquired channel, so you're right that failing probably makes sense there." ]

[ "untriaged" ]

[ "src/main/java/com/google/devtools/build/lib/remote/blobstore/http/HttpBlobStore.java" ]

[ "src/main/java/com/google/devtools/build/lib/remote/blobstore/http/HttpBlobStore.java" ]

[]

[ "timestamp(timedelta=102305.0, similarity=0.9405752619525384)" ]

[ "We already have an issue here: https://github.com/bazelbuild/bazel/issues/5807\r\nClosing this one due to duplication." ]

[ "Have you thought of using a `typedef`? I think that'd be more type-safe and more readable than a macro.", "Have you thought of using an unnamed namespace? That restricts visibility and the binary will not export these symbols.\r\n\r\n(See https://google.github.io/styleguide/cppguide.html#Unnamed_Namespaces_and_Static_Variables)", "What's the reason for using `explicit`?", "The error message is confusing, \"INPUT RUNFILES\" can be understood as \"a list of input runfiles\". What's the intended meaning? Could you clarify the message?", "I think the argument type does not need to be a const reference. Can you think of a more efficient approach?", "What does the rest of the loop do when `use_metadata` is false?", "What's the reason to store an entry when `target` is empty?", "Is this call necessary? Doesn't `ifstream`'s destructor close the file?", "String comparison should ignore casing.", "This call fails if the directory is not empty. Is that expected?", "Why can we ignore it?", "Have you considered using a foreach-loop?", "`CreateFileW` returns `INVALID_HANDLE_VALUE` on error. Please compare `h` to that instead of to `0`.", "type-safety: use `FALSE` `FALSE` (BOOL) instead of `false` (bool).", "Do you need to do this before every test, or would it suffice to export these envvars just once?", "Why is there one more symlink on Windows?", "You are calling `readlink $i` multiple times. Could you cache its result?", "why is there one more symlink on Windows?", "same here about `readlink $i`", "Good idea. Replaced it with typedef and moved it into the class definition.", "Good to know that, thanks!", "Done", "I copied from the Linux version build-runfiles.cc, but I think it's unnecessary, so removed.", "If `use_metadata` is false, it means every line in the manifest is a runfile entry, then the loop will read each of them and add them in the manifest map. If `use_metadata` is true, it will skip all lines with even line number, because they are metadata lines.\r\nSee https://github.com/bazelbuild/bazel/blob/master/src/main/tools/build-runfiles.cc#L144", "Bazel sometimes need to create empty files under the runfiles tree. For example, for python binary, `__init__.py` is needed under every directory. Storing an entry with an empty `target` indicates we need to create such a file when creating the runfiles tree.", "Indeed, removed.", "`AsAbsoluteWindowsPath` will convert the path to a normalized path, than means both of them are already in lower case.", "Here we are deleting a symlink, so it won't fail even the target directory is not empty.", "Because if the directory is not empty, it means it contains some symlinks already pointing to the correct targets (we just called `ScanTreeAndPrune`). Then this directory shouldn't be removed in the first place. It's better to check if the directory is empty before deleting it, but there's not simple kernel API for that.", "Good idea! Done.", "Done.", "Done", "Indeed, it's confusing. Updated.", "Right, once is enough.", "Because for shell binary, we build both `bin` and `bin.exe`, but on Linux we only build `bin`", "Yes, done.", "Answered in above comment.", "Done.", "Thanks. Please add this as a code comment.", "Thanks! Could you add this explanation to the code?", "Thanks! The convention in manpages is to put mandatory arguments in angle brackets, e.g. `<manifest_file>`. I think doing that here would increase readability.", "The name `use_metadata` confuses me, it suggests that the function should use (or ignore) the metadata. WDYT? If you agree, can you think of a better name?", "Could you add this as a comment?", "Could you add this to the code as a comment?", "Thanks, please add it as a comment here too.", "Indeed, `ignore_metadata` is better.", "Just realized `use_metadata` is also the flag name and used some other places in Java code. Let me instead add a comment here to clarify.", "Done.", "Done", "Done!", "Done", "Done" ]

[ "type: feature request", "P2", "platform: windows", "area-Windows", "team-OSS" ]

[ "src/main/cpp/util/BUILD", "src/main/java/com/google/devtools/build/lib/analysis/config/BuildConfiguration.java", "src/main/tools/BUILD", "src/main/tools/build-runfiles-windows.cc" ]

[ "src/main/cpp/util/BUILD", "src/main/java/com/google/devtools/build/lib/analysis/config/BuildConfiguration.java", "src/main/tools/BUILD", "src/main/tools/build-runfiles-windows.cc" ]

[ "src/test/py/bazel/runfiles_test.py", "src/test/shell/integration/BUILD", "src/test/shell/integration/runfiles_test.sh" ]

[ "timestamp(timedelta=29963.0, similarity=0.9261200312519905)" ]

[ "Why am I assigned to this @laszlocsomor?", "IIRC you updated Protobuf in the past. If so, you have domain expertise. Am I confusing it with gRPC though?", "I have, so have other people. I currently have multiple P1s I am working on and don't have time to update protobuf anytime soon. There's a detailed description on how to update protobuf here: https://github.com/bazelbuild/bazel/tree/master/third_party/protobuf", "Thanks for the info. Let me take a look.", "If there's a cherrypick request, please add it to the release thread (https://github.com/bazelbuild/bazel/issues/5059)." ]

[ "please put `load()` statements at the top of the file" ]

[ "P1", "type: process" ]

[ "WORKSPACE", "src/main/protobuf/BUILD", "third_party/protobuf/3.6.1/BUILD", "third_party/protobuf/BUILD", "third_party/protobuf/README.md" ]

[ "WORKSPACE", "src/main/protobuf/BUILD", "third_party/protobuf/3.6.1/BUILD", "third_party/protobuf/3.6.1/compiler_config_setting.bzl", "third_party/protobuf/BUILD", "third_party/protobuf/README.md" ]

[ "src/test/shell/bazel/testdata/embedded_tools_srcs_deps", "src/test/shell/testenv.sh" ]

[ "timestamp(timedelta=7101.0, similarity=0.8553772038536177)" ]

[ "Why am I assigned to this @laszlocsomor?", "IIRC you updated Protobuf in the past. If so, you have domain expertise. Am I confusing it with gRPC though?", "I have, so have other people. I currently have multiple P1s I am working on and don't have time to update protobuf anytime soon. There's a detailed description on how to update protobuf here: https://github.com/bazelbuild/bazel/tree/master/third_party/protobuf", "Thanks for the info. Let me take a look.", "If there's a cherrypick request, please add it to the release thread (https://github.com/bazelbuild/bazel/issues/5059)." ]

[]

[ "P1", "type: process" ]

[ "third_party/protobuf/README.md" ]

[ "third_party/protobuf/README.md" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8646177122156461)" ]

[ "I found the root cause: similarly to https://github.com/bazelbuild/bazel/issues/2576, it's the java log path. Turns out we must pass the path with escaped backslashes instead of forward slashes. Working on a fix.", "How come it worked in 0.17rc1 then?", "I don't know yet. What I do know is that (a) the fix solves this issue, and (b) the fix removes an exception from the `<output_base>/server/jvm.out` that I've been aware of for a long time, but which never seemed to result in any problems (until now?)." ]

[ "Isn't this quadratic? Mind you, not that the strings in question are very long (cue PATH_MAX), but if there is a simple way...", "You're fully right. Let me fix that.", "nit: `std::count` could be used to count these numbers, but I'm not at all sure if it would result in simpler code.", "nit: you overwrite it anyway, why zero it out?", "I'm not sure if a blackbelt in C++ would be able to do it using the standard library, but I sure don't know how. string splitting/joining is apparently not in the standard library and this doesn't require retreating to a monastery for years to make sense of, so go ahead.", "Thanks for the idea! I'd need to use `count_if` and I think the resulting code would be neither more readable nor more efficient.", "Thanks! I'm also not aware of a good library function for this.", "AFAIK this is how you reserve space in a string.", "That sentiment, I can heartily agree with! :)" ]

[]

[ "src/main/cpp/startup_options.cc", "src/main/cpp/util/path_windows.cc" ]

[ "src/main/cpp/startup_options.cc", "src/main/cpp/util/path_windows.cc" ]

[ "src/test/shell/integration/execution_phase_tests.sh", "src/test/shell/integration/server_logging_test.sh" ]

[ "timestamp(timedelta=0.0, similarity=0.9339910474624503)" ]

[]

[ "Just curious: Do relative paths still work when prefixed with `\\\\?\\` ?", "Sorry - ignore that. I somehow thought that this method can also produce relative paths (GitHub showed the diff in a strange way, until I refreshed the page and suddenly it's 10x as large)", "TIL: you can have a directory named with `...`\r\nThanks! ;)", "Thanks!\r\nFor completeness' sake, even though you can create such a directory (e.g. from MSYS: `mkdir '/c/...'`) it's not fully usable from Windows Explorer and other shell applications.\r\nFor example, you cannot `cd` into it but you can list its contents and access its files." ]

[ "platform: windows", "area-Windows", "team-OSS" ]

[ "src/main/cpp/util/path_platform.h", "src/main/cpp/util/path_windows.cc" ]

[ "src/main/cpp/util/path_platform.h", "src/main/cpp/util/path_windows.cc" ]

[ "src/test/cpp/util/file_windows_test.cc", "src/test/cpp/util/path_windows_test.cc" ]

[ "timestamp(timedelta=86914.0, similarity=0.8436393502010401)" ]

[ "Thanks for bringing this to my attention. We're actually in the middle of a documentation audit this week, and this is exactly the type of issue we want to flag." ]

[]

[ "type: documentation (cleanup)", "P2", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/platform/ToolchainInfoApi.java" ]

[ "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/platform/ToolchainInfoApi.java" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.8645373485183019)" ]

[ "Thanks for bringing this to my attention. We're actually in the middle of a documentation audit this week, and this is exactly the type of issue we want to flag." ]

[]

[ "type: documentation (cleanup)", "P2", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/platform/PlatformCommonApi.java" ]

[ "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/platform/PlatformCommonApi.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9159548986968176)" ]

[ "Just for completeness, which NDK revision are you on?", "@jin r17b" ]

[]

[ "team-Android", "untriaged" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r10e/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r11/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r12/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r13/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r15/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r17/ArmCrosstools.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r10e/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r11/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r12/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r13/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r15/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r17/ArmCrosstools.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9232423870207832)" ]

[]

[]

[ "type: bug", "platform: linux" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/python/python_stub_template.txt" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/python/python_stub_template.txt" ]

[ "src/test/py/bazel/BUILD", "src/test/py/bazel/runfiles_sandboxed_test.py" ]

[ "timestamp(timedelta=0.0, similarity=0.8708933620718008)" ]

[ "`DEVELOPER_DIR` is also missing from the persistent worker env here", "Ok I think I discovered why this is the case. When building locally we hit this codepath: https://github.com/bazelbuild/bazel/blob/18bd7dec9609ff8924d9ebed5f14be523df4eec0/src/main/java/com/google/devtools/build/lib/sandbox/DarwinSandboxedSpawnRunner.java#L224-L225 which adds the variables we want from here https://github.com/bazelbuild/bazel/blob/1f684e1b87cd8881a0a4b33e86ba66743e32d674/src/main/java/com/google/devtools/build/lib/exec/apple/XcodeLocalEnvProvider.java#L66-L108 but persistent workers don't call this local env rewrite logic https://github.com/bazelbuild/bazel/blob/5a1ed1f24e74f8a608ee145133f399c1f34a8390/src/main/java/com/google/devtools/build/lib/worker/WorkerSpawnRunner.java#L131 meaning they don't add the variables we were hoping for.\r\n\r\nI'm not really sure what an acceptable solution for this is since I don't think normally we'd want to pass persistent workers the local environment logic, but in this case it's required for quite a few arguments.", "@allevato assuming you have some context on these, any thoughts?", "I've run into the same problem in my worker prototype, and my solution was to hack it by having my wrapper set DEVELOPER_DIR and SDKROOT by querying xcode-select/xcrun manually if they're not already set. Not ideal.", "Sounds like a problem with persistent workers. @keith wanna send us a patch? :-) cc: @philwo ", "@buchgr I'm happy to submit that fix, but is that a reasonable one? Just based on the naming tying the local environment objects to the persistent workers sounds like they're intentionally attached", "I've submitted a PR that fixes it in this way, please let me know what you think! https://github.com/bazelbuild/bazel/pull/6287" ]

[ "I've seen some different things passed for `/tmp` here, I assume there's something other than this to be more platform agnostic that I should pass?" ]

[]

[ "src/main/java/com/google/devtools/build/lib/worker/WorkerModule.java", "src/main/java/com/google/devtools/build/lib/worker/WorkerSpawnRunner.java" ]

[ "src/main/java/com/google/devtools/build/lib/worker/WorkerModule.java", "src/main/java/com/google/devtools/build/lib/worker/WorkerSpawnRunner.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9999999999999998)" ]

[]

[ "Can you use CppFileTypes.C_SOURCE + CPP_SOURCE?", "That means packing an additional dependency into bazel_tools which makes the binary bigger. I copied those values to Constants.CC_EXTENSIONS and use it here to check if the file is a cc file." ]

[ "P2", "coverage", "team-Rules-CPP" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/sh/BazelShTestRule.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/sh/ShBinary.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/sh/BazelShTestRule.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/sh/ShBinary.java" ]

[ "src/test/shell/bazel/bazel_coverage_test.sh", "tools/test/CoverageOutputGenerator/java/com/google/devtools/coverageoutputgenerator/Constants.java", "tools/test/CoverageOutputGenerator/java/com/google/devtools/coverageoutputgenerator/Coverage.java", "tools/test/CoverageOutputGenerator/java/com/google/devtools/coverageoutputgenerator/Main.java", "tools/test/CoverageOutputGenerator/javatests/com/google/devtools/coverageoutputgenerator/CoverageTest.java" ]

[ "timestamp(timedelta=21605.0, similarity=0.9246448239924371)" ]

[ "Sounds like a great idea!", "Just hit this too. It looks like the fix is trivial (adding a no-op `doc` parameter), so I've mailed https://github.com/bazelbuild/bazel/pull/6936 . Note that rules using this param won't work with any older Bazel versions." ]

[]

[ "type: feature request", "P2", "team-ExternalDeps" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/repository/skylark/SkylarkRepositoryModule.java", "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/repository/RepositoryModuleApi.java", "src/main/java/com/google/devtools/build/skydoc/fakebuildapi/repository/FakeRepositoryModule.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/repository/skylark/SkylarkRepositoryModule.java", "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/repository/RepositoryModuleApi.java", "src/main/java/com/google/devtools/build/skydoc/fakebuildapi/repository/FakeRepositoryModule.java" ]

[ "src/test/java/com/google/devtools/build/skydoc/testdata/misc_apis_test/input.bzl" ]

[ "timestamp(timedelta=1.0, similarity=0.9242516503873975)" ]

[ "Sounds like a great idea!", "Just hit this too. It looks like the fix is trivial (adding a no-op `doc` parameter), so I've mailed https://github.com/bazelbuild/bazel/pull/6936 . Note that rules using this param won't work with any older Bazel versions." ]

[]

[ "type: feature request", "P2", "team-ExternalDeps" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/repository/skylark/SkylarkRepositoryModule.java", "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/repository/RepositoryModuleApi.java", "src/main/java/com/google/devtools/build/skydoc/fakebuildapi/repository/FakeRepositoryModule.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/repository/skylark/SkylarkRepositoryModule.java", "src/main/java/com/google/devtools/build/lib/skylarkbuildapi/repository/RepositoryModuleApi.java", "src/main/java/com/google/devtools/build/skydoc/fakebuildapi/repository/FakeRepositoryModule.java" ]

[ "src/test/java/com/google/devtools/build/skydoc/testdata/misc_apis_test/input.bzl" ]

[ "timestamp(timedelta=65146.0, similarity=0.9342796586228654)" ]

[ "r19 beta 2 is available, r19 is expected to ship soon.\r\n\r\nhttps://developer.android.com/ndk/downloads/", "NDK 19 has dropped: https://developer.android.com/ndk/downloads/#stable-downloads", "Link to changelog: https://github.com/android-ndk/ndk/wiki/Changelog-r19", "After speaking with @hlopko, the right way forward here is to implement the NDK CROSSTOOL in Starlark. This isn't a necessary blocker for NDK 19, since we can adapt the NDK 18 configuration for 19, but it would be nice to turn down proto CROSSTOOL generation to move to the modern approach.\r\n\r\n@hlopko's suggestion:\r\n\r\n> write a rule ndk_cc_toolchain_config that has attributes for things that are different between versions, and only instantiate that in BUILD files.\r\n\r\nMore information: \r\n\r\n- https://github.com/bazelbuild/bazel/issues/7320\r\n- https://docs.bazel.build/versions/master/cc-toolchain-config-reference.html\r\n", "@jin do you plan on supporting `lld` too ?", "@steeve definitely. We want the Android rules to be 100% compatible with the recommended set of tools in the NDK.", "The NDK CROSSTOOL has been ported to Starlark with massive help from @hlopko - thanks!\r\n\r\nGoing through the [changelog](https://github.com/android-ndk/ndk/wiki/Changelog-r19):\r\n\r\n> # Announcements\r\n\r\n> Developers should begin testing their apps with LLD. AOSP has switched to using LLD by default and the NDK will use it by default in the next release. BFD and Gold will be removed once LLD has been through a release cycle with no major unresolved issues (estimated r21). Test LLD in your app by passing -fuse-ld=lld when linking.\r\n \r\n> Note: lld does not currently support compressed symbols on Windows. See Issue 888. Clang also cannot generate compressed symbols on Windows, but this can be a problem when using artifacts built from Darwin or Linux.\r\n\r\nWe will need to add support for this.\r\n\r\n> The Play Store will require 64-bit support when uploading an APK beginning in August 2019. Start porting now to avoid surprises when the time comes. For more information, see this blog post.\r\n\r\nWe may want to add a warning message about this when building for non-64-bit.\r\n\r\n> Issue 780: Standalone toolchains are now unnecessary. Clang, binutils, the sysroot, and other toolchain pieces are now all installed to $NDK/toolchains/llvm/prebuilt/<host-tag> and Clang will automatically find them. Instead of creating a standalone toolchain for API 26 ARM, instead invoke the compiler directly from the NDK:\r\n> \r\n> $ $NDK/toolchains/llvm/prebuilt/<host-tag>/bin/armv7a-linux-androideabi26-clang++ src.cpp\r\n> For r19 the toolchain is also installed to the old path to give build systems a chance to adapt to the new layout. The old paths will be removed in r20.\r\n\r\nWe must definitely support this, but the effects of it should be transparent to end users. It's not clear what the impact of having a standalone toolchain will be on the Starlark crosstool configuration.\r\n \r\n> The make_standalone_toolchain.py script will not be removed. It is now unnecessary and will emit a warning with the above information, but the script will remain to preserve existing workflows.\r\n\r\nNot applicable, we don't use this.\r\n\r\n> If you're using ndk-build, CMake, or a standalone toolchain, there should be no change to your workflow. This change is meaningful for maintainers of third-party build systems, who should now be able to delete some Android-specific code. For more information, see the Build System Maintainers guide.\r\n\r\nMore reading: https://android.googlesource.com/platform/ndk/+/ndk-release-r19/docs/BuildSystemMaintainers.md\r\n\r\n> ndk-depends has been removed. We believe that ReLinker is a better solution to native library loading issues on old Android versions.\r\n\r\nNot applicable, we don't use this.\r\n\r\n> Issue 862: The GCC wrapper scripts which redirected to Clang have been removed, as they are not functional enough to be drop in replacements.\r\n\r\nThis should be addressed by the standalone toolchain support.\r\n\r\n---\r\n\r\n> r19c\r\n> ----\r\n> \r\n> * [Issue 912]: Prevent the CMake toolchain file from clobbering a user\r\n> specified `CMAKE_FIND_ROOT_PATH`.\r\n> * [Issue 920]: Fix clang wrapper scripts on Windows.\r\n> \r\n> [Issue 912]: https://github.com/android-ndk/ndk/issues/912\r\n> [Issue 920]: https://github.com/android-ndk/ndk/issues/920\r\n> \r\n> r19b\r\n> ----\r\n> \r\n> * [Issue 855]: ndk-build automatically disables multithreaded linking for LLD\r\n> on Windows, where it may hang. It is not possible for the NDK to detect this\r\n> situation for CMake, so CMake users and custom build systems must pass\r\n> `-Wl,--no-threads` when linking with LLD on Windows.\r\n\r\nWe'll need to take note of this for the `lld` support.\r\n\r\n> * [Issue 849]: Fixed unused command line argument warning when using standalone\r\n> toolchains to compile C code.\r\n> * [Issue 890]: Fixed `CMAKE_FIND_ROOT_PATH`. CMake projects will no longer\r\n> search the host's sysroot for headers and libraries.\r\n> * [Issue 906]: Explicitly set `-march=armv7-a` for 32-bit ARM to workaround\r\n> Clang not setting that flag automatically when using `-fno-integrated-as`.\r\n> This fix only affects ndk-build and CMake. Standalone toolchains and custom\r\n> build systems will need to apply this fix themselves.\r\n> * [Issue 907]: Fixed `find_path` for NDK headers in CMake.\r\n> \r\n> [Issue 849]: https://github.com/android-ndk/ndk/issues/849\r\n> [Issue 890]: https://github.com/android-ndk/ndk/issues/890\r\n> [Issue 907]: https://github.com/android-ndk/ndk/issues/907\r\n> \r\n> Changes\r\n> -------\r\n> \r\n> * Updated Clang to r339409.\r\n> * C++ compilation now defaults to C++14.\r\n\r\nNeed to update default std to `-std=c++14`.\r\n\r\n> * [Issue 780]: A complete NDK toolchain is now installed to the Clang\r\n> directory. See the announcements section for more information.\r\n> * ndk-build no longer removes artifacts from `NDK_LIBS_OUT` for ABIs not\r\n> present in `APP_ABI`. This enables workflows like the following:\r\n> \r\n> ```bash\r\n> for abi in armeabi-v7a arm64-v8a x86 x86_64; do\r\n> ndk-build APP_ABI=$abi\r\n> done\r\n> ```\r\n> \r\n> Prior to this change, the above workflow would remove the previously built\r\n> ABI's artifacts on each successive build, resulting in only x86_64 being\r\n> present at the end of the loop.\r\n> * ndk-stack has been rewritten in Python.\r\n> * [Issue 776]: To better support LLD, ndk-build and CMake no longer pass\r\n> `-Wl,--fix-cortex-a8` by default.\r\n> * CPUs that require this fix are uncommon in the NDK's supported API range\r\n> (16+).\r\n> * If you need to continue supporting these devices, add\r\n> `-Wl,--fix-cortex-a8` to your `APP_LDFLAGS` or `CMAKE_C_FLAGS`, but note\r\n> that LLD will not be adding support for this workaround.\r\n> * Alternatively, use the Play Console to [blacklist] Cortex-A8 CPUs to\r\n> disallow your app from being installed on those devices.\r\n\r\nOur ARM crosstool add this `--fix-cortex-a8` linker flag, so we should drop them here.\r\n\r\n```\r\nsrc/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/ArmCrosstools.java\r\n149: .addLinkerFlag(\"-Wl,--fix-cortex-a8\")\r\n```\r\n\r\n> * [Issue 798]: The ndk-build and CMake options to disable RelRO and noexecstack\r\n> are now ignored. All code is built with RelRO and non-executable stacks.\r\n\r\nNot sure how/if we do this, will need to investigate.\r\n\r\n> * [Issue 294]: All code is now built with a subset of [compiler-rt] to provide\r\n> a complete set of compiler built-ins for Clang.\r\n> \r\n> [Issue 294]: https://github.com/android-ndk/ndk/issues/294\r\n> [Issue 776]: https://github.com/android-ndk/ndk/issues/776\r\n> [Issue 798]: https://github.com/android-ndk/ndk/issues/798\r\n> [blacklist]: https://support.google.com/googleplay/android-developer/answer/7353455?hl=en\r\n> [compiler-rt]: https://compiler-rt.llvm.org/\r\n> \r\n> Known Issues\r\n> ------------\r\n> \r\n> * This is not intended to be a comprehensive list of all outstanding bugs.\r\n> * [Issue 888]: lld does not support compressed symbols on Windows. Clang also\r\n> cannot generate compressed symbols on Windows, but this can be a problem when\r\n> using artifacts built from Darwin or Linux.\r\n> * [Issue 360]: `thread_local` variables with non-trivial destructors will cause\r\n> segfaults if the containing library is `dlclose`ed on devices running M or\r\n> newer, or devices before M when using a static STL. The simple workaround is\r\n> to not call `dlclose`.\r\n> * [Issue 70838247]: Gold emits broken debug information for AArch64. AArch64\r\n> still uses BFD by default.\r\n> * [Issue 855]: LLD may hang on Windows when using multithreaded linking.\r\n> ndk-build will automatically disable multithreaded linking in this situation,\r\n> but CMake users and custom build systems should pass `-Wl,--no-threads` when\r\n> using LLD on Windows. The other linkers and operating systems are unaffected.\r\n\r\nAddressed above.\r\n\r\n> * [Issue 884]: Third-party build systems must pass `-fno-addrsig` to Clang for\r\n> compatibility with binutils. ndk-build, CMake, and standalone toolchains\r\n> handle this automatically.\r\n\r\nLooks like standalone toolchains support this, but will be good to verify.\r\n\r\n> * [Issue 906]: Clang does not pass `-march=armv7-a` to the assembler when using\r\n> `-fno-integrated-as`. This results in the assembler generating ARMv5\r\n> instructions. Note that by default Clang uses the integrated assembler which\r\n> does not have this problem. To workaround this issue, explicitly use\r\n> `-march=armv7-a` when building for 32-bit ARM with the non-integrated\r\n> assembler, or use the integrated assembler. ndk-build and CMake already\r\n> contain these workarounds.\r\n\r\nWe'll need to add this workaround.\r\n\r\n> * This version of the NDK is incompatible with the Android Gradle plugin\r\n> version 3.0 or older. If you see an error like\r\n> `No toolchains found in the NDK toolchains folder for ABI with prefix: mips64el-linux-android`,\r\n> update your project file to [use plugin version 3.1 or newer]. You will also\r\n> need to upgrade to Android Studio 3.1 or newer.\r\n> \r\n> [Issue 360]: https://github.com/android-ndk/ndk/issues/360\r\n> [Issue 70838247]: https://issuetracker.google.com/70838247\r\n> [Issue 855]: https://github.com/android-ndk/ndk/issues/855\r\n> [Issue 884]: https://github.com/android-ndk/ndk/issues/884\r\n> [Issue 888]: https://github.com/android-ndk/ndk/issues/888\r\n> [Issue 906]: https://github.com/android-ndk/ndk/issues/906\r\n> [use plugin version 3.1 or newer]: https://developer.android.com/studio/releases/gradle-plugin#updating-plugin\r\n\r\n\r\n", "Addressed in https://github.com/bazelbuild/bazel/pull/8524" ]

[]

[ "type: feature request", "P1", "team-Android" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/AndroidNdkCrosstoolsR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/ApiLevelR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/NdkMajorRevisionR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/X86Crosstools.java" ]

[ "src/test/java/com/google/devtools/build/lib/bazel/rules/android/AndroidNdkRepositoryTest.java", "src/test/shell/bazel/android/android_ndk_integration_test.sh" ]

[ "timestamp(timedelta=641.0, similarity=0.9231411341872593)" ]

[ "r19 beta 2 is available, r19 is expected to ship soon.\r\n\r\nhttps://developer.android.com/ndk/downloads/", "NDK 19 has dropped: https://developer.android.com/ndk/downloads/#stable-downloads", "Link to changelog: https://github.com/android-ndk/ndk/wiki/Changelog-r19", "After speaking with @hlopko, the right way forward here is to implement the NDK CROSSTOOL in Starlark. This isn't a necessary blocker for NDK 19, since we can adapt the NDK 18 configuration for 19, but it would be nice to turn down proto CROSSTOOL generation to move to the modern approach.\r\n\r\n@hlopko's suggestion:\r\n\r\n> write a rule ndk_cc_toolchain_config that has attributes for things that are different between versions, and only instantiate that in BUILD files.\r\n\r\nMore information: \r\n\r\n- https://github.com/bazelbuild/bazel/issues/7320\r\n- https://docs.bazel.build/versions/master/cc-toolchain-config-reference.html\r\n", "@jin do you plan on supporting `lld` too ?", "@steeve definitely. We want the Android rules to be 100% compatible with the recommended set of tools in the NDK.", "The NDK CROSSTOOL has been ported to Starlark with massive help from @hlopko - thanks!\r\n\r\nGoing through the [changelog](https://github.com/android-ndk/ndk/wiki/Changelog-r19):\r\n\r\n> # Announcements\r\n\r\n> Developers should begin testing their apps with LLD. AOSP has switched to using LLD by default and the NDK will use it by default in the next release. BFD and Gold will be removed once LLD has been through a release cycle with no major unresolved issues (estimated r21). Test LLD in your app by passing -fuse-ld=lld when linking.\r\n \r\n> Note: lld does not currently support compressed symbols on Windows. See Issue 888. Clang also cannot generate compressed symbols on Windows, but this can be a problem when using artifacts built from Darwin or Linux.\r\n\r\nWe will need to add support for this.\r\n\r\n> The Play Store will require 64-bit support when uploading an APK beginning in August 2019. Start porting now to avoid surprises when the time comes. For more information, see this blog post.\r\n\r\nWe may want to add a warning message about this when building for non-64-bit.\r\n\r\n> Issue 780: Standalone toolchains are now unnecessary. Clang, binutils, the sysroot, and other toolchain pieces are now all installed to $NDK/toolchains/llvm/prebuilt/<host-tag> and Clang will automatically find them. Instead of creating a standalone toolchain for API 26 ARM, instead invoke the compiler directly from the NDK:\r\n> \r\n> $ $NDK/toolchains/llvm/prebuilt/<host-tag>/bin/armv7a-linux-androideabi26-clang++ src.cpp\r\n> For r19 the toolchain is also installed to the old path to give build systems a chance to adapt to the new layout. The old paths will be removed in r20.\r\n\r\nWe must definitely support this, but the effects of it should be transparent to end users. It's not clear what the impact of having a standalone toolchain will be on the Starlark crosstool configuration.\r\n \r\n> The make_standalone_toolchain.py script will not be removed. It is now unnecessary and will emit a warning with the above information, but the script will remain to preserve existing workflows.\r\n\r\nNot applicable, we don't use this.\r\n\r\n> If you're using ndk-build, CMake, or a standalone toolchain, there should be no change to your workflow. This change is meaningful for maintainers of third-party build systems, who should now be able to delete some Android-specific code. For more information, see the Build System Maintainers guide.\r\n\r\nMore reading: https://android.googlesource.com/platform/ndk/+/ndk-release-r19/docs/BuildSystemMaintainers.md\r\n\r\n> ndk-depends has been removed. We believe that ReLinker is a better solution to native library loading issues on old Android versions.\r\n\r\nNot applicable, we don't use this.\r\n\r\n> Issue 862: The GCC wrapper scripts which redirected to Clang have been removed, as they are not functional enough to be drop in replacements.\r\n\r\nThis should be addressed by the standalone toolchain support.\r\n\r\n---\r\n\r\n> r19c\r\n> ----\r\n> \r\n> * [Issue 912]: Prevent the CMake toolchain file from clobbering a user\r\n> specified `CMAKE_FIND_ROOT_PATH`.\r\n> * [Issue 920]: Fix clang wrapper scripts on Windows.\r\n> \r\n> [Issue 912]: https://github.com/android-ndk/ndk/issues/912\r\n> [Issue 920]: https://github.com/android-ndk/ndk/issues/920\r\n> \r\n> r19b\r\n> ----\r\n> \r\n> * [Issue 855]: ndk-build automatically disables multithreaded linking for LLD\r\n> on Windows, where it may hang. It is not possible for the NDK to detect this\r\n> situation for CMake, so CMake users and custom build systems must pass\r\n> `-Wl,--no-threads` when linking with LLD on Windows.\r\n\r\nWe'll need to take note of this for the `lld` support.\r\n\r\n> * [Issue 849]: Fixed unused command line argument warning when using standalone\r\n> toolchains to compile C code.\r\n> * [Issue 890]: Fixed `CMAKE_FIND_ROOT_PATH`. CMake projects will no longer\r\n> search the host's sysroot for headers and libraries.\r\n> * [Issue 906]: Explicitly set `-march=armv7-a` for 32-bit ARM to workaround\r\n> Clang not setting that flag automatically when using `-fno-integrated-as`.\r\n> This fix only affects ndk-build and CMake. Standalone toolchains and custom\r\n> build systems will need to apply this fix themselves.\r\n> * [Issue 907]: Fixed `find_path` for NDK headers in CMake.\r\n> \r\n> [Issue 849]: https://github.com/android-ndk/ndk/issues/849\r\n> [Issue 890]: https://github.com/android-ndk/ndk/issues/890\r\n> [Issue 907]: https://github.com/android-ndk/ndk/issues/907\r\n> \r\n> Changes\r\n> -------\r\n> \r\n> * Updated Clang to r339409.\r\n> * C++ compilation now defaults to C++14.\r\n\r\nNeed to update default std to `-std=c++14`.\r\n\r\n> * [Issue 780]: A complete NDK toolchain is now installed to the Clang\r\n> directory. See the announcements section for more information.\r\n> * ndk-build no longer removes artifacts from `NDK_LIBS_OUT` for ABIs not\r\n> present in `APP_ABI`. This enables workflows like the following:\r\n> \r\n> ```bash\r\n> for abi in armeabi-v7a arm64-v8a x86 x86_64; do\r\n> ndk-build APP_ABI=$abi\r\n> done\r\n> ```\r\n> \r\n> Prior to this change, the above workflow would remove the previously built\r\n> ABI's artifacts on each successive build, resulting in only x86_64 being\r\n> present at the end of the loop.\r\n> * ndk-stack has been rewritten in Python.\r\n> * [Issue 776]: To better support LLD, ndk-build and CMake no longer pass\r\n> `-Wl,--fix-cortex-a8` by default.\r\n> * CPUs that require this fix are uncommon in the NDK's supported API range\r\n> (16+).\r\n> * If you need to continue supporting these devices, add\r\n> `-Wl,--fix-cortex-a8` to your `APP_LDFLAGS` or `CMAKE_C_FLAGS`, but note\r\n> that LLD will not be adding support for this workaround.\r\n> * Alternatively, use the Play Console to [blacklist] Cortex-A8 CPUs to\r\n> disallow your app from being installed on those devices.\r\n\r\nOur ARM crosstool add this `--fix-cortex-a8` linker flag, so we should drop them here.\r\n\r\n```\r\nsrc/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r18/ArmCrosstools.java\r\n149: .addLinkerFlag(\"-Wl,--fix-cortex-a8\")\r\n```\r\n\r\n> * [Issue 798]: The ndk-build and CMake options to disable RelRO and noexecstack\r\n> are now ignored. All code is built with RelRO and non-executable stacks.\r\n\r\nNot sure how/if we do this, will need to investigate.\r\n\r\n> * [Issue 294]: All code is now built with a subset of [compiler-rt] to provide\r\n> a complete set of compiler built-ins for Clang.\r\n> \r\n> [Issue 294]: https://github.com/android-ndk/ndk/issues/294\r\n> [Issue 776]: https://github.com/android-ndk/ndk/issues/776\r\n> [Issue 798]: https://github.com/android-ndk/ndk/issues/798\r\n> [blacklist]: https://support.google.com/googleplay/android-developer/answer/7353455?hl=en\r\n> [compiler-rt]: https://compiler-rt.llvm.org/\r\n> \r\n> Known Issues\r\n> ------------\r\n> \r\n> * This is not intended to be a comprehensive list of all outstanding bugs.\r\n> * [Issue 888]: lld does not support compressed symbols on Windows. Clang also\r\n> cannot generate compressed symbols on Windows, but this can be a problem when\r\n> using artifacts built from Darwin or Linux.\r\n> * [Issue 360]: `thread_local` variables with non-trivial destructors will cause\r\n> segfaults if the containing library is `dlclose`ed on devices running M or\r\n> newer, or devices before M when using a static STL. The simple workaround is\r\n> to not call `dlclose`.\r\n> * [Issue 70838247]: Gold emits broken debug information for AArch64. AArch64\r\n> still uses BFD by default.\r\n> * [Issue 855]: LLD may hang on Windows when using multithreaded linking.\r\n> ndk-build will automatically disable multithreaded linking in this situation,\r\n> but CMake users and custom build systems should pass `-Wl,--no-threads` when\r\n> using LLD on Windows. The other linkers and operating systems are unaffected.\r\n\r\nAddressed above.\r\n\r\n> * [Issue 884]: Third-party build systems must pass `-fno-addrsig` to Clang for\r\n> compatibility with binutils. ndk-build, CMake, and standalone toolchains\r\n> handle this automatically.\r\n\r\nLooks like standalone toolchains support this, but will be good to verify.\r\n\r\n> * [Issue 906]: Clang does not pass `-march=armv7-a` to the assembler when using\r\n> `-fno-integrated-as`. This results in the assembler generating ARMv5\r\n> instructions. Note that by default Clang uses the integrated assembler which\r\n> does not have this problem. To workaround this issue, explicitly use\r\n> `-march=armv7-a` when building for 32-bit ARM with the non-integrated\r\n> assembler, or use the integrated assembler. ndk-build and CMake already\r\n> contain these workarounds.\r\n\r\nWe'll need to add this workaround.\r\n\r\n> * This version of the NDK is incompatible with the Android Gradle plugin\r\n> version 3.0 or older. If you see an error like\r\n> `No toolchains found in the NDK toolchains folder for ABI with prefix: mips64el-linux-android`,\r\n> update your project file to [use plugin version 3.1 or newer]. You will also\r\n> need to upgrade to Android Studio 3.1 or newer.\r\n> \r\n> [Issue 360]: https://github.com/android-ndk/ndk/issues/360\r\n> [Issue 70838247]: https://issuetracker.google.com/70838247\r\n> [Issue 855]: https://github.com/android-ndk/ndk/issues/855\r\n> [Issue 884]: https://github.com/android-ndk/ndk/issues/884\r\n> [Issue 888]: https://github.com/android-ndk/ndk/issues/888\r\n> [Issue 906]: https://github.com/android-ndk/ndk/issues/906\r\n> [use plugin version 3.1 or newer]: https://developer.android.com/studio/releases/gradle-plugin#updating-plugin\r\n\r\n\r\n", "Addressed in https://github.com/bazelbuild/bazel/pull/8524" ]

[]

[ "type: feature request", "P1", "team-Android" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java" ]

[ "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/AndroidNdkCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/AndroidNdkCrosstoolsR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/ApiLevelR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/ArmCrosstools.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/NdkMajorRevisionR19.java", "src/main/java/com/google/devtools/build/lib/bazel/rules/android/ndkcrosstools/r19/X86Crosstools.java" ]

[ "src/test/java/com/google/devtools/build/lib/bazel/rules/android/AndroidNdkRepositoryTest.java" ]

[ "timestamp(timedelta=0.0, similarity=0.9904610207971085)" ]

[ "Release blocker for 0.18 #5963 " ]

[ "the legacy behavior would be to have this hidden behind --master_blazerc, in case that matters to you at this point" ]

[ "P0" ]

[ "src/main/cpp/option_processor.cc" ]

[ "src/main/cpp/option_processor.cc" ]

[ "src/test/cpp/rc_file_test.cc" ]

[ "timestamp(timedelta=0.0, similarity=0.8946550413544082)" ]

[ "This can also be reproduced easily on a Linux system directly:\r\n\r\n- Check out the Bazel source code.\r\n- Change the line endings of `tools/genrule/genrule-setup.sh` to CRLF (e.g., using `:set ff=dos` in Vim and saving the file).\r\n- Run it directly using Bash. It will print the same errors as shown above." ]

[]

[ "P1", "area-Windows", "untriaged", "team-OSS" ]

[ ".gitattributes" ]

[ ".gitattributes" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.873985520432397)" ]

[ "@laszlocsomor there should be only one team label. Is that better for your team or for External dependencies?", "I cannot tell. We have the Windows expertise, the Ext.Deps team has the rule logic expertise.", "I spent some time with #7063 so I'm also investigating this bug now.", "More observation:\r\n- When I add a new file to the external repo, Bazel *does* notice this, as proven by logging I added. Bazel creates a new `RegularDirectoryListingValue` SkyValue, however the downstream action is not re-executed.\r\n- If I re-run the build once more without any change, Bazel creates yet another `RegularDirectoryListingValue` SkyValue, whose hashcode is different from that in the previous bullet point, and this time Bazel also re-executes the downstream genrule.", "Found the culprit. It's complex, so I'll write it down in a document and paste here later.", "## Repro\r\n\r\nLet's have two directories, \"A\" and \"B\". \"A\" contains a WORKSPACE and BUILD file, while \"B\" contains just a dummy file:\r\n\r\n/src/A/WORKSPACE:\r\n```\r\nnew_local_repository(\r\n name = \"B\",\r\n build_file_content = \"\"\"\r\nfilegroup(\r\n name = \"F\",\r\n srcs = glob([\"*.txt\"]),\r\n visibility = [\"//visibility:public\"],\r\n)\r\n\"\"\",\r\n path = \"/src/B\",\r\n)\r\n```\r\n\r\n/src/A/BUILD:\r\n```\r\ngenrule(\r\n name = \"G\",\r\n srcs = [\"@B//:F\"],\r\n outs = [\"g.txt\"],\r\n cmd = \"echo $(SRCS) > $@\",\r\n)\r\n```\r\n\r\n/src/B/a.txt:\r\n```\r\ndummy\r\n```\r\n\r\n## Bug\r\n\r\nIf we build \"//:G\" inside /src/A, then g.txt correctly lists external/B/a.txt and nothing else.\r\n\r\nIf we add a new file /src/B/b.txt and rebuild :G, then g.txt should list external/B/a.txt and external/B/b.txt -- and it does on Linux but not on Windows.\r\n\r\nIf we build a third time on Windows without doing anything, g.txt will contain the correct output.\r\n\r\n## Analysis\r\n\r\nnew_local_repository creates symlinks to every file and directory in its root, e.g. external/B/a.txt is a symlink to /src/B/a.txt. This is done as part of computing the REPOSITORY_DIRECTORY SkyValue, which is done by evaluating the [RepositoryDelegatorFunction](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryDelegatorFunction.java#L179), which calls into [NewLocalRepositoryFunction.fetch](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/rules/repository/NewLocalRepositoryFunction.java#L100), which creates these symlinks. On Linux these are all symlinks indeed, but on Windows only the directory symlinks are really symlinks, the files are copies (because we cannot create file symlinks).\r\n\r\nOn both Linux and Windows, FILE_STATEs and DIRECTORY_LISTING_STATEs of paths under the external directory have special handling -- they all depend on the corresponding REPOSITORY_DIRECTORY. This dependency is established [here](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/FileStateFunction.java#L48) and [here](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/DirectoryListingStateFunction.java#L54), by the [ExternalFilesHelper.maybeHandleExternalFile()](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/ExternalFilesHelper.java#L208) --&gt; [RepositoryFunction.addExternalFilesDependencies()](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryFunction.java#L510) call chain, and it makes perfect sense, because it is the repository's SkyFunction that populates the external/B directory and creates the symlinks (or files) under it. Furthermore, the REPOSITORY_DIRECTORY depends on the DIRECTORY_LISTING_STATE of /src/B, so on both platforms Skyframe recomputes the REPOSITORY_DIRECTORY after we added /src/B/b.txt and correctly creates the new symlink (copy) under external/B/. This is all good.\r\n\r\nThe difference is that on Linux the GLOB depends on external/B/a.txt while on Windows it doesn't, because on Linux it's a symlink but on Windows it's a regular file. The reason is a performance optimization: GLOB only depends on FILE_STATEs of symlinks and directories but not files, to avoid recomputing the glob when a matched file's contents change, but to dirty the glob when a directory (or symlink to one) is replaced by a file with the same name (which affects recursive globbing).\r\n\r\nOn both Linux and Windows, the GLOB depends on the DIRECTORY_LISTING_STATE of external/B, and on Linux the GLOB also depends on the FILE_STATE of external/B/a.txt (because it's a symlink).\r\nOn both platforms, the FILE_STATE of external/B/a.txt depends on the REPOSITORY_DIRECTORY of @B, because the file is under external/. When we add the new file to B and rebuild, Skyframe's invalidation phase dirties the DIRECTORY_LISTING_STATE of /src/B, which dirties the REPOSITORY_DIRECTORY, which dirties the FILE_STATE of external/B/a.txt and on Linux this also dirties the GLOB.\r\n\r\nThis is a big difference: on Linux the GLOB is dirty but on Windows it's not. Upon closer inspection though, we notice that on Linux the GLOB is not recomputed but revalidated in the second build, and its SkyValue does not contain the newly added file -- however, the PACKAGE for \"@B//\" is recomputed and it does have the new glob contents and there's a new TARGET for @B//:b.txt -- what's going on?!\r\n\r\nTurns out there's some package reloading logic [in PackageFunction](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/PackageFunction.java#L1188) that [uses Skyframe's dirtiness information](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/PackageFunction.java#L908) to decide which GLOBs to recompute, but doesn't actually recompute the GlobValue. On Linux the GLOB is dirtied in the second build but on Windows it's not, so on Linux we re-glob (and find external/B/b.txt) but on Windows we don't (even though external/B/b.txt is there).\r\n\r\nWhy doesn't the GLOB depend on the DIRECTORY_LISTING_STATE of external/B though? [It actually does](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/skyframe/GlobFunction.java#L135). So why isn't this DIRECTORY_LISTING_STATE dirtied on Windows? It actually is! The problem is that the DIRECTORY_LISTING_STATE of external/B does not depend on the REPOSITORY_DIRECTORY of @B, because the [RepositoryFunction.addExternalFilesDependencies() call doesn't add this dependency if the path is the same as the repository's root path, in order to avoid a dependency cycle](https://github.com/bazelbuild/bazel/blob/d3d90630fa53d7d99e7e3de0a88bc0dfebefd50d/src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryFunction.java#L534-L551). Which is cool for the FILE_STATE of /src/B, but unnecessary for DIRECTORY_LISTING_STATE of it -- that's the bug.\r\n\r\nIf we change RepositoryFunction.addExternalFilesDependencies() to depend on the REPOSITORY_DIRECTORY for DIRECTORY_LISTING_STATEs, even of external/B itself, then everything works correctly on Windows. Becaues then adding the new file to /src/B will dirty the DIRECTORY_LISTING_STATE of /src/B, dirtying the REPOSITORY_DIRECTORY, dirtying the DIRECTORY_LISTING_STATE of /external/B, dirtying the GLOB, so the partial package reloading logic will re-compute that GLOB, even on Windows.\r\n\r\nIronically the bug was there on Linux too, just never manifested… because on Linux, external/B always contains symlinks, so the GLOB depends on their FILE_STATEs, and thus depends on the REPOSITORY_DIRECTORY." ]

[]

[ "type: bug", "P1", "area-Windows", "team-OSS" ]

[ "src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/DirectoryListingStateFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/ExternalFilesHelper.java", "src/main/java/com/google/devtools/build/lib/skyframe/FileStateFunction.java" ]

[ "src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/DirectoryListingStateFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/ExternalFilesHelper.java", "src/main/java/com/google/devtools/build/lib/skyframe/FileStateFunction.java" ]

[ "src/test/shell/bazel/BUILD", "src/test/shell/bazel/new_local_repo_test.sh" ]

[ "timestamp(timedelta=0.0, similarity=0.8539077854347784)" ]

[ "Can you provide example output showing where the message should be?", "This is a snippet from an output from a failed build for TF:\r\n\r\nERROR: /tmpfs/tmp/bazel/external/org_tensorflow/third_party/toolchains/preconfig/ubuntu14.04/cuda9.0-cudnn7/cuda/BUILD:206:1: Couldn't build file external/org_tensorflow/third_party/toolchains/preconfig/ubuntu14.04/cuda9.0-cudnn7/cuda/cuda/include/CL/cl.h: Executing genrule @org_tensorflow//third_party/toolchains/preconfig/ubuntu14.04/cuda9.0-cudnn7/cuda:cuda-include failed (Exit 1)\r\ncp: cannot stat '/usr/local/cuda-9.0/include/CL/cl.h': No such file or directory\r\n[9,866 / 10,412] 278 / 503 tests, 278 failed; no action\r\n\r\nThe error message coming from the use of --verbose_failures is:\r\n\"cp: cannot stat '/usr/local/cuda-9.0/include/CL/cl.h': No such file or directory\"\r\n\r\nThis error occurred because the wrong platform was selected to run this action (due to a user mistake in flags)\r\n\r\nIt would have been ideal to get something like\r\n\r\ncp: cannot stat '/usr/local/cuda-9.0/include/CL/cl.h': No such file or directory\r\nPlatform info: this action ran in <something about the platform that either shows all details here or points me to the platform target that was selected for this action>\r\n\r\nPlease let me know if you want more details / a more comprehensive example.\r\n" ]

[ "nit: Bogus indentation, here and below.", "Why null? Why cannot we use the same value as in the branch below?", "Why all these no-op changes here? I mean, I guess they're fine, but they make it harder to see what actually changed :)", "Sorry, I just wanted to shorten my new target, but my regex was a little greedy and fixed all of the local targets. I can revert everything except command-utils if you'd like.", "In an ideal world, we'd have two different execution platforms here:\r\n- The actual spawn's execution platform, which would describe the sandbox (and ideally give configuration for it)\r\n- The sandbox's execution platform, which describes the environment that the sandbox container can run in.\r\n\r\nWe don't actually define either of these very well right now, but it's definitely wrong to assume they're the same, so it's best to just pass null and not describe it.", "The formatter was confused because this is correct but previous commits left the indent wrong on the previous lines. Fixed it.", "(I see now that the two branches are subtly different in what they call `get*` on.)\r\n\r\nBut can the platforms between the sandbox and the action ever be different?", "One of my goals is to better couple execution strategies and platforms. Ideally, we'd end up with platform definitions like this (this is entirely a strawman, don't take any actual syntax seriously):\r\n\r\n```\r\nplatform(name = \"linux_sandbox\",\r\n constraint_values = [...],\r\n execution_strategy = \"linux-sandbox\",\r\n execution_properties = \"\"\"\r\n sandbox_base_dir: /tmp/sandbox\r\n sandbox_allow_networking: false\r\n \"\"\")\r\n\r\nplatform(name = \"host\",\r\n constraint_values = [...],\r\n)\r\n```\r\n\r\nIn this case, both of these are execution platforms, but \"host\" is the platform the sandbox runs on. They might have identical constraint_values, but they mean very different things.", "Oh man, I was going to point you at someone else about this platform stuff... until I dug who \"katre\" was, so I cannot point you at yourself ;-)" ]

[ "type: feature request", "P1", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/BUILD", "src/main/java/com/google/devtools/build/lib/actions/BUILD", "src/main/java/com/google/devtools/build/lib/exec/AbstractSpawnStrategy.java", "src/main/java/com/google/devtools/build/lib/query2/BUILD", "src/main/java/com/google/devtools/build/lib/sandbox/AbstractSandboxSpawnRunner.java", "src/main/java/com/google/devtools/build/lib/sandbox/BUILD", "src/main/java/com/google/devtools/build/lib/util/CommandFailureUtils.java", "src/main/java/com/google/devtools/build/lib/util/CommandUtils.java" ]

[ "src/main/java/com/google/devtools/build/lib/BUILD", "src/main/java/com/google/devtools/build/lib/actions/BUILD", "src/main/java/com/google/devtools/build/lib/exec/AbstractSpawnStrategy.java", "src/main/java/com/google/devtools/build/lib/query2/BUILD", "src/main/java/com/google/devtools/build/lib/sandbox/AbstractSandboxSpawnRunner.java", "src/main/java/com/google/devtools/build/lib/sandbox/BUILD", "src/main/java/com/google/devtools/build/lib/util/CommandFailureUtils.java", "src/main/java/com/google/devtools/build/lib/util/CommandUtils.java" ]

[ "src/test/java/com/google/devtools/build/lib/BUILD", "src/test/java/com/google/devtools/build/lib/util/CommandFailureUtilsTest.java" ]

[ "timestamp(timedelta=1.0, similarity=0.8519683283943104)" ]

[]

[]

[ "type: bug", "P2", "category: misc > testing", "area-EngProd", "team-OSS" ]

[ ".bazelci/postsubmit.yml", ".bazelci/presubmit.yml" ]

[ ".bazelci/postsubmit.yml", ".bazelci/presubmit.yml" ]

[]

[ "timestamp(timedelta=2.0, similarity=0.8870653956623342)" ]

[ "I think such PR will be welcome - any objections @philwo ?", "That would be great, yes. Thanks!" ]

[]

[ "type: feature request", "P2", "area-EngProd", "team-OSS" ]

[ "scripts/packages/debian/BUILD", "scripts/packages/debian/control" ]

[ "scripts/packages/debian/BUILD", "scripts/packages/debian/control" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8461699383865314)" ]

[ "Confirmed the behavior you're seeing.\r\n\r\nFrom what I can see from the review history, it looks like there was never any design intention to exclude config_setting from aliases. Pinging @lberki for confirmation or rebuttal.\r\n\r\nAs to your deeper goal, that's an interesting use of the feature. It seems like another variant of [OR chaining](https://docs.bazel.build/versions/master/configurable-attributes.html#or-chaining). Do you see this as bringing novel functionality even beyond that?", "Thanks for the reply.\r\n\r\nOR chaining would work, but what I'm trying to achieve is to define the OR in only one place, and have other places use the result of it. I guess you could probably define a constant or macro and then reuse it, but that's awkward compared to the below, which allows consumers to consider it as any other `config_setting`.\r\n\r\nThe concrete example I added was this:\r\n```python\r\nalias(\r\n name = \"modules_enabled\",\r\n actual = select({\r\n \":java9\": \":java9\",\r\n \":java11\": \":java11\",\r\n # rhs just needs to be anything that won't match\r\n # :java11 will not match if we get to here\r\n \"//conditions:default\": \":java11\",\r\n }),\r\n)\r\n```\r\n\r\nWe have Java runtimes/toolchains controlled by a `--define` flag. Sometimes libraries need to add module-related `javacopts`; this allows them to do:\r\n```python\r\njava_library(\r\n name = \"lib\",\r\n javacopts = select({\r\n \"//tools/java:modules_enabled\": [\"--add-exports=...\"],\r\n \"//conditions:default\": [],\r\n }),\r\n)\r\n```\r\nThis is the cleanest solution I found for this issue. The `alias` itself is kinda hacky as I mentioned, but it's worth it so that all consumers have a single intuitive flag to switch on for this behavior.\r\n\r\nI could be missing a better answer, which I'd be happy to adopt.", "Without engaging with the meat of the bug report, whether aliasing works is different for each rule you mentioned above:\r\n * `config_setting` should just work, I think it's a documentation bug\r\n * `package_group` may or may not work -- they are regrettably a little different from regular configured targets, so I wouldn't be surprised if there is a bug lurking somewhere that prevents them from working with `alias`, but I see no fundamental reason for them not to. I'd prefer them to be regular rules since they are special mostly for historical reason and then this wouldn't even be a question.\r\n* `test_suite` is special-cased in a few places, so I'd be even less surprised if `alias` did not work for them than for `package_group`. It should, though.", "This approach doesn't look bad to me at all, and I think probably fits your use case as good as you're going to get.\r\n\r\nRe: the documentation: what wording would you find clearer given the the comment above? Happy to patch or take a patch from you.", "Sorry, was that last question for me? I'm not sure I have enough context to answer -- I don't know what constitutes a 'regular' rule.\r\n\r\nFrom a little more experimentation, neither `package_group` nor `test_suite` may be aliased. However:\r\n- `package_group` is listed as a function, not a rule, so I'm not sure why there's the expectation that it would work in the first place.\r\n- `test_suite` doesn't error or anything, it just doesn't run the tests. That seems to be covered by the next part of the doc that says \"Tests are not run if their alias is mentioned on the command line.\"\r\n\r\nSo maybe the whole line is redundant and could be removed?", "Yeah, removing the line sounds easiest. Will do..." ]

[ "Are you sure this works for `package_group` and `test_suite`? Is this functionality tested? `test_suite` rules are (unfortunately) special-cased in a number of ways e.g. in `bazel query` and other places expanding them (unfortunately, we have more than one) and I have no idea whether aliases are taken into account there.", "For `package_group` and `test_suite` I'm following the thought process described at https://github.com/bazelbuild/bazel/issues/6449#issuecomment-433712899.", "I think you are technically correct, but the fact that `package_group` is not a proper rule is a very obscure piece of information (except for us who have worked on Bazel since forever) and `test_suite` *is* a rule. Why not call this out explicitly?\r\n\r\n(btw, `package_group` totally should be a rule. Totally my fault that it's not.)", "Call out what explicitly? Like, don't get into the details that `package_group` and `test_suite` are \"special\" but call them out.. how?\r\n\r\nI don't mind adding some simple tests to check that alias works in a standard build use, but I wouldn't be enthused to check all the corner cases with query and who knows where else. :p", "Updated to just remove reference to `config_setting`, which is really the point of the bug. There's no evidence the comments on `package_group` and `test_suite` are confusing people, so let's just leave them be." ]

[ "type: documentation (cleanup)", "P3", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/rules/Alias.java" ]

[ "src/main/java/com/google/devtools/build/lib/rules/Alias.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8592612044934141)" ]

[ "This is a bad first-time experience. See https://github.com/bazelbuild/bazel/issues/5254", "@laszlocsomor \r\n\r\nI get the same error: `Bazel on Windows requires MSYS2 Bash, but we could not find it` for every command using the binaries for Bazel provided by the `@bazel/bazel` v0.20.0 npm package.\r\n\r\nSo the only way to run Bazel on Windows is to install MSYS2 bash and disable the path conversion?\r\n\r\nThe doc says to prefer Command Prompt or PowerShell: https://docs.bazel.build/versions/master/windows.html#running-bazel-msys2-shell-vs-command-prompt-vs-powershell", "@robisim74 : try Bazel 0.21, that should no longer complain about a missing MSYS2 Bash. Please let me know if you still find problems with it!", "Ok, thanks! I will try to install v0.21 (still not published on npm) and let you know.", "@laszlocsomor I tried to download version 0.21 directly on my machine in `c:\\bazel`, and run `C:\\> .\\bazel help`, but nothing: same error.\r\n\r\nFor the moment, I installed MSYS2 bash to work.", "@robisim74 : You're right. I never merged the corresponding PR (https://github.com/bazelbuild/bazel/pull/6582) because it broke some tests.\r\n\r\nInstalling MSYS2 is the best \"workaround\" for now. Sorry about that.", "Bazel 0.23 still complains about missing MSYS2:\r\n```\r\nC:\\src\\tmp>c:\\Users\\laszlocsomor\\Downloads\\bazel-0.23.0-windows-x86_64.exe info release\r\nERROR: bash.exe not found on PATH\r\nBazel on Windows requires MSYS2 Bash, but we could not find it.\r\nIf you do not have it installed, you can install MSYS2 from\r\n http://repo.msys2.org/distrib/msys2-x86_64-latest.exe\r\n\r\nIf you already have it installed but Bazel cannot find it,\r\nset BAZEL_SH environment variable to its location:\r\n set BAZEL_SH=c:\\path\\to\\msys2\\usr\\bin\\bash.exe\r\n[bazel ERROR src/main/cpp/blaze_util_windows.cc:1462] bash.exe not found on PATH\r\n[bazel INFO src/main/cpp/blaze_util_windows.cc:1477] BAZEL_SH detection took 0 msec, found\r\n```" ]

[]

[ "type: bug", "P2", "area-Windows", "team-OSS" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/blaze_util_platform.h", "src/main/cpp/blaze_util_windows.cc" ]

[ "src/main/cpp/blaze.cc", "src/main/cpp/blaze_util_platform.h", "src/main/cpp/blaze_util_windows.cc" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.843525691914465)" ]

[ "This is a bad first-time experience. See https://github.com/bazelbuild/bazel/issues/5254" ]

[]

[ "type: bug", "P3", "area-Windows", "team-OSS" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[ "src/test/py/bazel/BUILD", "src/test/py/bazel/first_time_use_test.py" ]

[ "timestamp(timedelta=25713.0, similarity=0.851142177143935)" ]

[ "cc @katre, since this is new in `0.19` (#6116), but probably not a release blocker.", "This is a small enough use-case that I don't want to risk a breakage due to cherrypick. We should try and address it for 0.20, however.", "I'm getting the same error message for **non-symlinks** on Windows with Bazel 0.19.0:\r\n```\r\nD:\\src\\codebase>bazel build //...\r\nWARNING: The following rc files are no longer being read, please transfer their contents or import their path into one of the standard rc files:\r\nd:\\src\\codebase/.bazelrc\r\n```", "Same as @baskus. Can somebody please confirm this is not the intended behaviour?", "Assigning to @dslomov to decide if this is a blocker for 0.20", "/sub, this is causing confusing issues on the Mac too. The system_rc is `/etc/bazel.bazelrc` on the Mac, but `/etc` is a symlink to `/private/etc`. This results in confusing warning messages like:\r\n\r\n```\r\n$ cat /etc/bazel.bazelrc\r\nbuild --jobs=4\r\n$ bazel version\r\nWARNING: The following rc files are no longer being read, please transfer\r\ntheir contents or import their path into one of the standard rc files:\r\n/etc/bazel.bazelrc\r\n```", "Thanks @excitoon for linking to #6138! That might indeed be the same issue.\r\n\r\n~Someone with a Mac at hand, could you please Bazel built at https://github.com/bazelbuild/bazel/commit/e9a908560133770614eca89ef64681bdf3f04b3e and see if that commit fixes this issue?~ Oh wait, never mind, it happens on Linux too I see.", "Confirming that https://github.com/bazelbuild/bazel/commit/e9a908560133770614eca89ef64681bdf3f04b3e fixes this issue too, so I'll say this is a duplicate of https://github.com/bazelbuild/bazel/issues/6138." ]

[ "nit: not that this is very important, but can you maybe get away with `$TEST_TMPDIR` instead? It's kinda confusing because `$HOME` usually means, well, the home directory of some user.", "nit: DedupeBazelRcPaths? (it's not a getter)\r\n\r\nAlso, why \"Existing\"? It doesn't check existence, does it?", "On Linux we do the same: https://github.com/bazelbuild/bazel/blob/420b9533534951b82c1e13e6d83fa67390e95778/src/main/cpp/blaze_util_posix.cc#L205\r\n\r\nBazel actually sets a mock $HOME for the test.", "The canonical path can only be computed of existing paths, so the paths indeed exist.\r\n\r\nRenamed the method to its old name." ]

[ "untriaged", "team-Bazel" ]

[ "src/main/cpp/blaze_util.cc", "src/main/cpp/blaze_util.h", "src/main/cpp/blaze_util_windows.cc", "src/main/cpp/option_processor-internal.h", "src/main/cpp/option_processor.cc", "src/main/cpp/rc_file.cc", "src/main/cpp/rc_file.h", "src/main/cpp/startup_options.cc" ]

[ "src/main/cpp/blaze_util.cc", "src/main/cpp/blaze_util.h", "src/main/cpp/blaze_util_windows.cc", "src/main/cpp/option_processor-internal.h", "src/main/cpp/option_processor.cc", "src/main/cpp/rc_file.cc", "src/main/cpp/rc_file.h", "src/main/cpp/startup_options.cc" ]

[ "src/test/cpp/rc_file_test.cc" ]

[ "timestamp(timedelta=1.0, similarity=0.9130243255272305)" ]

[ "This is a bad first-time user experience: MSYS is required just to run `bazel info`. (And also to build C++ without any genrule dependencies.) See https://github.com/bazelbuild/bazel/issues/5254", "The reason I care about this scenario, even though the repro requires commenting out code, is because I'd like to fix https://github.com/bazelbuild/bazel/issues/6462 by not erroring out if Bash is missing.", "Understand, I will fix this." ]

[ "2018?", "don't you have to enable extensions to use `::` for comments?", "stale comment?", "`exit /b 1` ?", "Flipping the branches to get `if [condition] then [action] else [alternative]` is easier to read than `if not [condition] then [alternative] else [action]`.", "Done", "No, `::` is just like `#` in python", "Thanks!", "Sounds good", "Indeed!" ]

[ "type: bug", "area-Windows", "team-OSS" ]

[ "tools/cpp/vc_installation_error.bat.tpl", "tools/cpp/windows_cc_configure.bzl" ]

[ "tools/cpp/msys_gcc_installation_error.bat", "tools/cpp/vc_installation_error.bat.tpl", "tools/cpp/windows_cc_configure.bzl" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.9244399942700263)" ]

[ "A small enhancement: discovering the issue(s) the contributor's commit(s) closes, and linking to those issues too.\r\n\r\ne.g.\r\n\r\n```\r\nThis release contains contributions from many people at Google, as well as:\r\n\r\n* Jane Doe: #1234 Fixes foo bug, #1235 Added bar feature\r\n* John Doe: #1236 Improved baz performance\r\n```", "@jin Currently we're not linking individual contributors to single issues, just listing everyone who contributed. I'm not sure trying to do that is going to be useful in the release notes: anyone who is interested can check the Git history themselves." ]

[]

[ "type: feature request", "P2", "area-EngProd", "team-OSS" ]

[ "scripts/release/relnotes.sh" ]

[ "scripts/release/relnotes.sh" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.944254643173867)" ]

[ "@spomorski, we have https://docs.bazel.build/versions/master/be/common-definitions.html, where should the `exec_compatible_with` attribute be documented? It can currently be used with any test rule, and shell rule, or genrule, but we may expand that set in the future.", "I think that page is fine. You didn't mention build rules, does it work\nwith just test/shell/genrules (I might just be misunderstanding\nshell/genrules)? If it works for build rules, let's just stick it in\n\"Attributes common to all build rules.\"\n\nOn Tue, Oct 30, 2018 at 3:24 PM katre <[email protected]> wrote:\n\n> @spomorski <https://github.com/spomorski>, we have\n> https://docs.bazel.build/versions/master/be/common-definitions.html,\n> where should the exec_compatible_with attribute be documented? It can\n> currently be used with any test rule, and shell rule, or genrule, but we\n> may expand that set in the future.\n>\n> —\n> You are receiving this because you were mentioned.\n> Reply to this email directly, view it on GitHub\n> <https://github.com/bazelbuild/bazel/issues/6558#issuecomment-434434630>,\n> or mute the thread\n> <https://github.com/notifications/unsubscribe-auth/AbdI8hfZz5YpyVNoQ1Ya7B5uzkbzbDRCks5uqKd7gaJpZM4YCrC7>\n> .\n>\n", "Sorry, by shell rules I mean sh_library, sh_binary, and sh_test.\r\n\r\nI'll start adding some documentation to the common definitions." ]

[ "\"The\" -> \"A\" and mention they're not the only ones. How about:\r\n\r\n> A list of <...>constraint_values</...> that must be present in the execution platform for this target. This is in addition to any constraints already imposed by the rule type. Constraints are used to restrict the list of available execution platforms during <...>toolchain resolution</...>.", "I'd remove \"being configured\", it's clear from context what target we're talking about. Even rule authors don't always know exactly what's entailed by the configuration phase or think in terms of \"configured targets\".", "\"that do not match these constraints\" -- Arguably it's not just \"the target's\" constraints if we're talking about constraints specified by the rule as well.", "Done.", "Done.", "Done." ]

[ "type: documentation (cleanup)", "P1", "team-Configurability" ]

[ "site/docs/platforms.md", "site/docs/toolchains.md", "src/main/java/com/google/devtools/build/docgen/PredefinedAttributes.java" ]

[ "site/docs/platforms.md", "site/docs/toolchains.md", "src/main/java/com/google/devtools/build/docgen/PredefinedAttributes.java", "src/main/java/com/google/devtools/build/docgen/templates/attributes/common/exec_compatible_with.html" ]

[]

[ "timestamp(timedelta=6128.0, similarity=0.8487044611062099)" ]

[ "Thanks for the detailed bug report! :) I agree that this should be cleaned up.\r\nI think your suggestion to guard this #define with an #ifndef block is fine.\r\n\r\nI'll fix this right now." ]

[]

[ "type: bug", "P2", "category: misc > misc", "area-EngProd", "team-OSS" ]

[ "src/tools/singlejar/transient_bytes.h" ]

[ "src/tools/singlejar/transient_bytes.h" ]

[]

[ "timestamp(timedelta=46216.0, similarity=0.8406726706695553)" ]

[ "Cc @lberki @iirina ", "Related: #6661.", "@Monnoroch There's no requirement anymore to have the same embedded JDK version as we use for the host javabase so #6661 is talking about updating the JDK version, it's a different JDK.\r\n\r\nIndependently, we also plan to upgrade the embedded JDK after we have submitted some other changes and hopefully it'll land before we cut 0.22.\r\n\r\nIf I understand the new java container support correctly, you probably still want to set a JVM flag in the container. The JVM uses by default 1/4 of the physical available memory for the heap. In a container, you may have less memory available than what's physically in your machine. Now, if the heap grows too large, Bazel gets killed. With the new container support, the \"physical available memory\" is limited to the memory available in the container. Thus, the heap will not grow too large, but it's using 1/4 of the available memory. That may not be enough and you might want to set MaxRamFraction in your CI environment.\r\n", "> Independently, we also plan to upgrade the embedded JDK after we have submitted some other changes and hopefully it'll land before we cut 0.22.\r\n\r\nI tried 0.22rc2 yesterday and it is still using JDK 9.", "David, that's correct. I wasn't able to properly test and land the changes mostly because of holidays. I'll definitely do it before 0.23.", "@meisterT Thanks. Let me know, if you have a commit somewhere. I would be happy to build Bazel from scratch and test it against Gerrit Code Review and related projects." ]

[]

[ "type: feature request", "P2", "area-EngProd", "team-Rules-Java", "team-OSS" ]

[ "WORKSPACE" ]

[ "WORKSPACE" ]

[]

[ "timestamp(timedelta=73717.0, similarity=0.845320405934995)" ]

[ "After some debugging, the problem is much simpler: the `--extra_toolchains` flag was not allowing multiple uses, so the last value passed was replacing the previous ones.\r\n\r\nI'll send a patch to fix that, until that is released you can use the multiple value format:\r\n```\r\nbazel build --extra_toolchains=//toolchains/docker:default_linux_toolchain,//toolchains/docker:default_windows_toolchain,//toolchains/docker:default_osx_toolchain\r\n```\r\nOr, even use the target pattern version:\r\n```\r\nbazel build --extra_toolchains=//toolchains/docker:all\r\n```\r\n" ]

[]

[ "type: bug", "P2", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/ToolchainContext.java", "src/main/java/com/google/devtools/build/lib/analysis/ToolchainResolver.java", "src/main/java/com/google/devtools/build/lib/skyframe/ToolchainResolutionFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/ToolchainResolutionValue.java" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/ToolchainContext.java", "src/main/java/com/google/devtools/build/lib/analysis/ToolchainResolver.java", "src/main/java/com/google/devtools/build/lib/skyframe/ToolchainResolutionFunction.java", "src/main/java/com/google/devtools/build/lib/skyframe/ToolchainResolutionValue.java" ]

[ "src/test/java/com/google/devtools/build/lib/BUILD", "src/test/java/com/google/devtools/build/lib/analysis/ToolchainResolverTest.java", "src/test/java/com/google/devtools/build/lib/rules/platform/ToolchainTestCase.java", "src/test/java/com/google/devtools/build/lib/skyframe/RegisteredToolchainsFunctionTest.java", "src/test/java/com/google/devtools/build/lib/skyframe/ToolchainResolutionFunctionTest.java" ]

[ "timestamp(timedelta=0.0, similarity=0.916451556493324)" ]

[ "cc @lberki ", "Doc for java_import is also broken:\r\n\r\n0.20.0: https://docs.bazel.build/versions/0.20.0/be/java.html#java_import \r\n\r\njava_import(name, deps, data, compatible_with, constraints, deprecation, distribs, exports, features, jars, licenses, neverlink, proguard_specs, restricted_to, runtime_deps, srcjar, tags, testonly, visibility)\r\n\r\n0.21.0: https://docs.bazel.build/versions/0.21.0/be/java.html#java_import\r\n\r\njava_import(name, deps, data, compatible_with, deprecation, distribs, exports, features, licenses, restricted_to, runtime_deps, tags, testonly, visibility)", "cc @iirina @buchgr @fweikert ", "Please fix this, not finding this rule in the documentation cost me a considerable amount of time. :cry: \r\nWriting docs for implemented code is not all that that much work compared to writing the code in the first place.", "Bisected to https://github.com/bazelbuild/bazel/commit/c1d7672707a4d8898b0cd37b5695804ca047d0b8.\r\n\r\n```\r\njingwen@jingwen:/tmp\r\n$ curl http://localhost:12345/versions/master/be/overview.html | grep java_runtime\r\n % Total % Received % Xferd Average Speed Time Time Time Current\r\n Dload Upload Total Spent Left Speed\r\n 0 0 0 0 0 0 0 0 --:--100 37307 100 37307 0 0 2602k 0 --:--:-- --:--:-- --:--:-- 2802k\r\n```\r\n\r\nAt 1db9e66b3d3666eeed37db286ece0f227c45e47b, the commit before c1d7672707a4d8898b0cd37b5695804ca047d0b8\r\n\r\n```\r\njingwen@jingwen:/tmp\r\n$ curl http://localhost:12345/versions/master/be/overview.html | grep java_runtime\r\n % Total % Received % Xferd Average Speed Time Time Time Current\r\n Dload Upload Total Spent Left Speed\r\n 0 0 0 0 0 0 0 0 --:--100 37666 100 37666 0 0 2627k 0 --:--:- <a href=\" java.html#java_runtime\">\r\n- java_runtime\r\n--:--:-- --:--:-- 2627k\r\n```\r\n", "cc @cushon ", "Fix: https://github.com/bazelbuild/bazel/pull/7061" ]

[]

[ "type: documentation (cleanup)", "P1", "team-Rules-Java" ]

[ "src/main/java/com/google/devtools/build/lib/BUILD" ]

[ "src/main/java/com/google/devtools/build/lib/BUILD" ]

[]

[ "timestamp(timedelta=21022.0, similarity=0.8786289929358719)" ]

[ "@laurentlb, is this possibly related to ea4840042b0c087685ab4faffab58ba6906b06f5?", "Testing locally, the docs render correctly.", "Actually, no, they just break differently. Still investigating.", "There are two issues, I think.\r\n\r\n- The HTML is broken. I'm fixing it. With the fix, page should render correctly when testing locally.\r\n- Syntax highlighting is broken on the site since my update. I suspect we need to update Jekyll or Rouge (cc @philwo).", "Now that 4616fdd0e2f567ae2395803e4404054a79787b22 is in, should I close this, or keep it open and re-assign for the syntax highlighting part?", "Asking @laurentlb whether to keep this open or not.", "Note that all multi-line code blocks are also broken currently, on the website itself:\r\nhttps://docs.bazel.build/versions/master/skylark/aspects.html", "Documentation using multi-line code blocks are severely broken (https://docs.bazel.build/versions/master/android-instrumentation-test.html), I will submit a rollback to use redcarpet first. ", "The rollback (rollforward?) worked. Closing this for now." ]

[]

[ "type: bug", "type: documentation (cleanup)", "P1", "team-Bazel" ]

[ "tools/build_defs/pkg/README.md" ]

[ "tools/build_defs/pkg/README.md" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8540919667945694)" ]

[ "I also had this issue with Chrome. I have no idea why this error exists but you could just add in a DisplayName with the registry editor. I went into regedit to: `Computer\\HKEY_CURRENT_USER\\Software\\Microsoft\\Windows\\CurrentVersion\\Uninstall\\Google Chrome` and added in a new `String Value` and called it \"DisplayName\". That appears to silence the error.\r\n\r\nFor the sake of having logs, here's my log output with a similar error message for the DisplayName:\r\n\r\n```\r\nD:\\src>bazel version\r\nERROR: Failed to query DisplayName of HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Uninstall\\Google Chrome\r\nWARNING: --batch mode is deprecated. Please instead explicitly shut down your Bazel server using the command \"bazel shutdown\".\r\nINFO: Invocation ID: ee818f56-b420-4910-a9fe-b54b73650abe\r\nBuild label: 0.20.0\r\nBuild target: bazel-out/x64_windows-opt/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar\r\nBuild time: Fri Nov 30 14:38:51 2018 (1543588731)\r\nBuild timestamp: 1543588731\r\nBuild timestamp as int: 1543588731\r\n\r\n```", "Thanks @cnishina ! If I'm not mistaken the error is non-fatal. Is that correct?\r\nI'll fix the code not to show even the error. Bazel should just acknowledge that it can't read some registry key and carry on silently.\r\n\r\n(And for the record, that warning about \"--batch mode\" is also benign, it just indicates that you are not running Bazel from a workspace, i.e. `d:\\src` directory has no WORKSPACE file in it, nor do any of its parent directories.)", "That is correct, it is not a fatal error. I just commented on how I resolved the registry problem." ]

[]

[ "type: bug", "P3", "area-Windows", "team-OSS" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[ "src/main/cpp/blaze_util_windows.cc" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8694351320346887)" ]

[]

[ "This accesses the same information via a provider rather than via attribute (deprecated way to do that?).\r\nDocumentation says `default_runfiles` may return `None` so I've added a check.\r\nhttps://docs.bazel.build/versions/master/skylark/lib/DefaultInfo.html#default_runfiles", "Attribute `files_map` does not exist", "It seems there is no need to wrap the invocation in shell?", "These are default values", "Executable file is a single file", "This is the default" ]

[ "team-Rules-Server", "untriaged" ]

[ "tools/build_defs/pkg/pkg.bzl" ]

[ "tools/build_defs/pkg/pkg.bzl" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8793758775672843)" ]

[ "Thanks for the report!\r\n\r\nThe error comes from here:\r\nhttps://github.com/bazelbuild/bazel/blob/7fc967c4d6435de2bb4e34aac00ca2e499f55fca/src/tools/singlejar/mapped_file_windows.inc#L66\r\n\r\nThe caller is here:\r\nhttps://github.com/bazelbuild/bazel/blob/7fc967c4d6435de2bb4e34aac00ca2e499f55fca/src/tools/singlejar/output_jar.cc#L909\r\n\r\nThe comment above the failing line in `mapped_file_windows.inc` explains that empty files are not supported. I think the caller should handle empty files instead of attempting to map them.\r\n\r\nI'll try to fix this." ]

[]

[ "type: bug", "P2", "platform: windows", "area-Windows", "team-OSS" ]

[ "src/tools/singlejar/mapped_file_windows.inc", "src/tools/singlejar/output_jar_shell_test.sh" ]

[ "src/tools/singlejar/mapped_file_windows.inc", "src/tools/singlejar/output_jar_shell_test.sh" ]

[]

[ "timestamp(timedelta=169092.0, similarity=0.8650940885209347)" ]

[ "Closing this as a dupe of https://github.com/bazelbuild/bazel/issues/1707" ]

[]

[ "type: documentation (cleanup)", "P2", "team-Bazel" ]

[ "scripts/docs/doc_versions.bzl", "site/_config.yml" ]

[ "scripts/docs/doc_versions.bzl", "site/_config.yml" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.840598533354835)" ]

[]

[]

[ "type: documentation (cleanup)", "P3" ]

[ "src/main/java/com/google/devtools/build/docgen/skylark/SkylarkDoc.java" ]

[ "src/main/java/com/google/devtools/build/docgen/skylark/SkylarkDoc.java" ]

[ "src/test/java/com/google/devtools/build/docgen/SkylarkDocumentationTest.java" ]

[ "timestamp(timedelta=0.0, similarity=0.8635947801373915)" ]

[ "Workaround: call `bazel shutdown` then `bazel run //src:foo` again.", "I bet this is some symlink-related issue. I suspect Bazel on Windows copies files in the roots of external repositories instead of symlinking them, and that's why it doesn't notice the change.\r\n\r\nI'll proceed to check this theory.", "My theory is correct:\r\n- moving `gen_cc.bzl` to a subdirectory `sub` in under `repo1` will result in correct behavior, because `<output_base>/external/repo1/sub` is a junction pointing to the real `sub` directory\r\n- adding a file to or removing a file from the root of repo1 between builds is picked up, but modifying a file is not; to see this, check the contents of `<output_base>/external/repo1`\r\n- changing `new_local_repository` to `local_repository` and adding a BUILD file to `repo1` will result in correct behavior, even when the bzl file is in repo1's root; therefore this is a bug with `new_local_repository`, ~and this bug might actually be a duplicate of #3908~", "Could be the same bug as https://github.com/bazelbuild/bazel/issues/6351" ]

[]

[ "type: bug", "P1", "area-Windows", "team-OSS" ]

[ "src/main/java/com/google/devtools/build/lib/rules/repository/NewLocalRepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryDirectoryValue.java" ]

[ "src/main/java/com/google/devtools/build/lib/rules/repository/NewLocalRepositoryFunction.java", "src/main/java/com/google/devtools/build/lib/rules/repository/RepositoryDirectoryValue.java" ]

[ "src/test/py/bazel/bazel_external_repository_test.py" ]

[ "timestamp(timedelta=0.0, similarity=0.8615181094211928)" ]

[ "/sub", "Debugging so far (by running tests in parallel and debugging linux vs windows:\r\nThe location of the bazel_tools repo is different in the Windows tests: on Linux it is (correctly) bazel_tools_workspace (set via /workspace/WORKSPACE). On windows, the default value from /DEFAULT.WORKSPACE (which is /usr/local/.../embedded_tools) is being used.\r\n\r\nI can see that on Windows, the Package.Builder is seeing both versions of the rule added, but for some reason in this test the /workspace/WORKSPACE version isn't overwriting.\r\n\r\nMore debugging to come.", "Wow, that was painful.\r\n\r\nSource of the issue: the test sets OS to LINUX for purposes of the test, but the in-memory-filesystem checks the current value of OS to decide about things like drive strings and whether or not the filesystem is case-insensitive. If this is wrong, on Windows, when BazelAnalysisMock tries to re-write the WORKSPACE, it can't read the old value and instead overwrites it.\r\n\r\nI still don't know why that happens, though.", "Root cause: when the in-memory VFS wrote the file /workspace/WORKSPACE, the OS was Windows, so the written hash code was \"WORKSPACE\".toLowerCase().hashCode().\r\nWhen it re-wrote the WORKSPACE, the OS was Linux, so the hash code was \"WORKSPACE\".hashCode().\r\n\r\nThis lead to the effective WORKSPACE file being replaced, instead of appended, and so the test value of \"/bazel_tools_workspace\" for @bazel_tools was dropped.\r\n\r\nSheesh.", "Ugh.. Nice debugging @katre !" ]

[]

[ "type: bug", "P2", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/util/CPU.java", "src/main/java/com/google/devtools/build/lib/util/OS.java" ]

[ "src/main/java/com/google/devtools/build/lib/util/CPU.java", "src/main/java/com/google/devtools/build/lib/util/OS.java" ]

[ "src/test/java/com/google/devtools/build/lib/bazel/repository/BUILD", "src/test/java/com/google/devtools/build/lib/bazel/repository/LocalConfigPlatformFunctionTest.java" ]

[ "timestamp(timedelta=0.0, similarity=0.9200953409301609)" ]

[ "This flag is part of the work on #6849 .", "Changed the targetted flag-flip release, since this will require extensive testing and may need to wait for the projected migration plan for the cpu flag.", "Hey @katre, I just re-discovered this flag, what's needed to flip this? Or at least to get to a place where we can ask users to try and to re-enable it on the CI to see what's broken on downstream?", "Related: bazelbuild/rules_go#2089, bazelbuild/rules_cc#24, #8763, https://github.com/bazelbuild/bazel/issues/8766.", "Major work left:\r\n1. Set the right tags so that we get migration readiness reports from CI, which I apparently never get right.\r\n2. Write directions so that users who currently use cross-compilation with `--cpu` and `--host_cpu` and platform-related rules can add a platform mapping and continue work. I suspect this is a small set of users but I can't prove that.", "Looks like only Bazel is to be migrated: https://buildkite.com/bazel/bazelisk-plus-incompatible-flags/builds/175", "Thanks for the pointer!", "Logs: https://buildkite.com/bazel/bazelisk-plus-incompatible-flags/builds/191#annotation---incompatible_auto_configure_host_platform", "Since this flag is already flipped, I'm temporarily removing the `migration-0.29` label from it because https://github.com/bazelbuild/bazel/issues/9104 is preventing the incompatible flags pipeline to give correct result on Windows and our update to Bazelisk is not ready yet.\r\n\r\nFYI @dslomov " ]

[ "Is this line still within the char limit? Unsure whether this can be visualized in github review...", "One l too many in untill.", "Yes, 99 characters long according to vim.", "Fixed." ]

[ "P1", "team-Configurability", "incompatible-change" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/PlatformOptions.java" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/PlatformOptions.java" ]

[ "src/test/java/com/google/devtools/build/lib/rules/platform/PlatformTest.java", "src/test/java/com/google/devtools/build/lib/skyframe/PlatformMappingFunctionTest.java", "src/test/java/com/google/devtools/build/lib/skyframe/PlatformMappingValueTest.java", "src/test/java/com/google/devtools/build/lib/testutil/TestConstants.java", "src/test/shell/integration/java_integration_test.sh" ]

[ "timestamp(timedelta=0.0, similarity=0.9179705795408803)" ]

[ "We need to fix this for remote execution and caching across all protocols. In the case of remote execution, we'll also need to set `skip_cache_lookup` to `true` in order to ensure re-execution." ]

[ "So \"we don't cache failed actions\" is no longer correct for Bazel? If so, this comment is no longer true (even if Bazel won't cache failed actions itself).", "s/missed cache/cache miss/", "Why do we need to test for the metadata?", "why do you think the asserts below this line are valuable for the test?", "I think the cache and execution priority aren't important for this test, no?", "// Failed actions are treated as a cache miss mostly in order to avoid caching flaky actions (tests).", "Afaik these are all default values no?", "upload should be called because `uploadLocalResults` is `true`. Generally, if a failed action was cached which then succeeds locally we want to overwrite the failed result with the successful result in the cache. This would typically be used for flaky tests.", "Right, we don't. Cleaned up.", "We need to verify, that current action indeed treated as a cash miss => we do not attempt to download anything. \r\nAsserts in line 528-529 are obsolete, removing them.", "Whether or not something is being downloaded seems orthogonal to what should be tested here. I'd argue that all this test should verify is that it's indeed a cache miss. We then have a separate unit test that verifies what should/should not happen on a cache miss.", "`when(res.exitCode()).thenReturn(1);\r\n when(res.status()).thenReturn(Status.EXECUTION_FAILED);`\r\nbecause of these 2 lines the local action also failed => no upload expected to the cache.\r\n\r\nBut I think successful use case makes more sense here, so changed to cache miss + successful local execution => cache.upload(...)", "Fair point. Removed.", "nit: instead of mocking the spawn you can just use SpawnResult.Builder()" ]

[ "type: bug", "P1", "team-Remote-Exec" ]

[ "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnRunner.java" ]

[ "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnRunner.java" ]

[ "src/test/java/com/google/devtools/build/lib/remote/RemoteSpawnCacheTest.java", "src/test/java/com/google/devtools/build/lib/remote/RemoteSpawnRunnerTest.java" ]

[ "timestamp(timedelta=1.0, similarity=0.8952493440993834)" ]

[ "cc @meisterT " ]

[]

[ "P2", "area-EngProd", "team-Rules-Java", "team-OSS" ]

[ "scripts/packages/BUILD" ]

[ "scripts/packages/BUILD" ]

[]

[ "timestamp(timedelta=90479.0, similarity=0.8733058587681826)" ]

[ "Thanks for the report, I'll look into this today. It should definitely be a warning, not a hard crash.", "The actual cycle is this:\r\n- //tests:expansion_test\r\n- needs a toolchain of type @io_bazel_rules_m4//m4:toolchain_type\r\n- depends on all registered toolchains\r\n- includes @io_bazel_rules_m4//m4/toolchains:v1.4.18\r\n- depends on @io_bazel_rules_m4//m4/internal:deny_shell\r\n- needs a toolchain of type @bazel_tools//tools/cpp:toolchain_type\r\n- depends on all registered toolchains\r\n- cycle\r\n\r\nSo, as you said, the real error here is that RegisteredToolchainsFunction has a dependency on something that isn't providing ToolchainRuleInfo.\r\n\r\nNormally, RegisteredToolchainsFunction only accepts targets of a rule named \"toolchain\", \r\nwhich prevents this in the typical case, but still allows for the odd case where someone writes a Starlark rule named \"toolchain\" and uses that incorrectly as the argument to register_toolchains. This rule is applied when expanding target patterns, such as if you wrote `register_toolchains('//toolchains:all')`.\r\n\r\nThe reason you see this error is because RegisteredToolchainsFunction tries to be forgiving, so that if a target is directly specified as the argument to register_toolchains, it is kept even if the rule name is not \"toolchain\". The assumption here being that the user knows what they are doing and intended wrat they wrote.\r\n\r\nGiven this, the best way forward is to try to add better error handling in RegisteredToolchainsFunction, which I will look into adding.", "Reproduction:\r\n```sh\r\n$ git clone [email protected]:jmillikin/rules_m4.git\r\n$ cd rules_m4\r\n$ bazel build //tests:expansion_test --extra_toolchains=//m4/internal:toolchain_v1.4.18\r\n```\r\n\r\nOddly, a simpler reproduction test case I tried to wrote did show an error message:\r\n```\r\n$ bazel build --extra_toolchains=//:demo //:sample\r\nStarting local Bazel server and connecting to it...\r\nINFO: Invocation ID: 5958e443-ec4a-4054-8cc6-112d98bd94bf\r\nERROR: While resolving toolchains for target //:sample: invalid registered toolchain '//:demo': target does not provide the DeclaredToolchainInfo provider\r\nWARNING: errors encountered while analyzing target '//:sample': it will not be built\r\nINFO: Analysed target //:sample (6 packages loaded, 23 targets configured).\r\nINFO: Found 0 targets...\r\nERROR: command succeeded, but not all targets were analyzed\r\nINFO: Elapsed time: 2.956s, Critical Path: 0.02s\r\nINFO: 0 processes.\r\nFAILED: Build did NOT complete successfully\r\n```", "Okay, managed to get a simpler reproduction, because this relies on their being two toolchains, one of which depends on the other, and the top-level one being mis-configured in register_toolchains.\r\n\r\nLooking into implementing a cycle reporter to make the problem clear. I'm not sure there's any level of introspection in RegisteredToolchainsFunction that can catch this, due to the fact that the error happens while the toolchain targets are being loaded." ]

[]

[ "P2", "team-Configurability" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/ToolchainResolver.java" ]

[ "src/main/java/com/google/devtools/build/lib/analysis/ToolchainResolver.java" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8850649384482638)" ]

[ "For this flag and the one in #7197 - what if users don't want to download this archive, or want to change the location/manner in which this archive is obtained? This would be important for CI systems with no internet connectivity, and orgs that want to audit/control build-time dependencies (both in terms of artifacts and systems). Will there be a flag moving forward that will disable this download?", "@scottmin0r To not download the archive you have to set the flags `--java_toolchain` and `--host_java_toolchain` to point to local tools.", "Specifically what should I pass for those flag values? As per [this comment](https://github.com/bazelbuild/bazel/issues/6656#issuecomment-461084131) I've tried:\r\n\r\n(release 0.24.1)\r\n\r\n```\r\nbuild --javabase=@local_jdk//:jdk\r\nbuild --host_javabase=@local_jdk//:jdk\r\nbuild --java_toolchain=@bazel_tools//tools/jdk:toolchain_hostjdk8\r\nbuild --host_java_toolchain=@bazel_tools//tools/jdk:toolchain_hostjdk8\r\n```\r\n\r\nbut this still seems to try to download tools (with two different failure modes):\r\n\r\n```\r\nAnalyzing: 1166 targets (68 packages loaded, 2267 targets configured)\r\nAnalyzing: 1166 targets (68 packages loaded, 2267 targets configured)\r\nAnalyzing: 1166 targets (68 packages loaded, 2267 targets configured)\r\nAnalyzing: 1166 targets (68 packages loaded, 2267 targets configured)\r\nAnalyzing: 1166 targets (68 packages loaded, 2267 targets configured)\r\nERROR: /home/kbuilder/.cache/bazel/_bazel_kbuilder/94efa1614cde89691caae16b65396550/external/bazel_tools/tools/jdk/BUILD:218:1: no such package '@remote_java_tools//': java.io.IOException: Error downloading [https://mirror.bazel.build/bazel_java_tools/java_tools_pkg-0.5.1.tar.gz] to /home/kbuilder/.cache/bazel/_bazel_kbuilder/94efa1614cde89691caae16b65396550/external/remote_java_tools/java_tools_pkg-0.5.1.tar.gz: Network is unreachable (connect failed) and referenced by '@bazel_tools//tools/jdk:JacocoCoverageRunner'\r\n```\r\n\r\n```\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nAnalyzing: 56 targets (8 packages loaded, 587 targets configured)\r\nERROR: /home/kbuilder/.cache/bazel/_bazel_kbuilder/1858cc0ae663ea49d21c8c6b3132f559/external/bazel_tools/tools/jdk/BUILD:153:1: no such package '@remote_java_tools//': java.io.IOException: thread interrupted and referenced by '@bazel_tools//tools/jdk:genclass'\r\n```" ]

[]

[ "incompatible-change" ]

[ "src/main/java/com/google/devtools/build/lib/rules/java/JavaOptions.java" ]

[ "src/main/java/com/google/devtools/build/lib/rules/java/JavaOptions.java" ]

[ "src/test/java/com/google/devtools/build/lib/analysis/mock/BazelAnalysisMock.java", "src/test/java/com/google/devtools/build/lib/rules/java/JavaSkylarkApiTest.java" ]

[ "timestamp(timedelta=157133.0, similarity=0.8819002292370186)" ]

[]

[]

[ "incompatible-change" ]

[ "src/main/java/com/google/devtools/build/lib/rules/java/JavaOptions.java" ]

[ "src/main/java/com/google/devtools/build/lib/rules/java/JavaOptions.java" ]

[ "src/test/java/com/google/devtools/build/lib/analysis/mock/BazelAnalysisMock.java", "src/test/java/com/google/devtools/build/lib/rules/java/JavaSkylarkApiTest.java" ]

[ "timestamp(timedelta=153861.0, similarity=0.9533572855750213)" ]

[ "deprecated in https://github.com/bazelbuild/bazel/pull/7985 \r\nWill be removed in #7895 " ]

[]

[ "P2", "type: support / not a bug (process)", "team-Remote-Exec" ]

[ "src/main/java/com/google/devtools/build/lib/remote/options/RemoteOptions.java" ]

[ "src/main/java/com/google/devtools/build/lib/remote/options/RemoteOptions.java" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.9491581201419624)" ]

[ "cc @benjaminp " ]

[ "The javadoc needs a update.", "In the `RemoteSpawnCache`, the equivalent condition was moved before the concurrent modification check. It probably doesn't matter, but it doesn't hurt to be consistent either way.", "Is it worthwhile to keep the `containsExactly` by adding the `actionKey` to the list here?", "I'm wondering if there should be a helper for the above three lines. They appear a lot.", "done", "done", "I'd argue no, as this test checks that directory uploads work. it's a protocol detail that the action and command are also stored in the CAS that should be verified elsewhere.", "yep. done", "done" ]

[ "type: bug", "P2", "team-Remote-Exec" ]

[ "src/main/java/com/google/devtools/build/lib/remote/AbstractRemoteActionCache.java", "src/main/java/com/google/devtools/build/lib/remote/GrpcRemoteCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnRunner.java", "src/main/java/com/google/devtools/build/lib/remote/SimpleBlobStoreActionCache.java", "src/tools/remote/src/main/java/com/google/devtools/build/remote/worker/ExecutionServer.java" ]

[ "src/main/java/com/google/devtools/build/lib/remote/AbstractRemoteActionCache.java", "src/main/java/com/google/devtools/build/lib/remote/GrpcRemoteCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnCache.java", "src/main/java/com/google/devtools/build/lib/remote/RemoteSpawnRunner.java", "src/main/java/com/google/devtools/build/lib/remote/SimpleBlobStoreActionCache.java", "src/tools/remote/src/main/java/com/google/devtools/build/remote/worker/ExecutionServer.java" ]

[ "src/test/java/com/google/devtools/build/lib/remote/AbstractRemoteActionCacheTests.java", "src/test/java/com/google/devtools/build/lib/remote/GrpcRemoteCacheTest.java", "src/test/java/com/google/devtools/build/lib/remote/RemoteSpawnCacheTest.java", "src/test/java/com/google/devtools/build/lib/remote/RemoteSpawnRunnerTest.java", "src/test/java/com/google/devtools/build/lib/remote/SimpleBlobStoreActionCacheTest.java", "src/test/shell/bazel/remote/remote_execution_test.sh" ]

[ "timestamp(timedelta=1760.0, similarity=0.8561444596451517)" ]

[ "I need this for properly fixing #6400 (support for grpc.enable_http_proxy channel arg was added in gRPC 1.14).", "@tetromino this is about the gRPC Java dependency. I believe you need https://github.com/bazelbuild/bazel/issues/2804", "The Google auth library in third_party/auth is quite old. It did not work with my company's proxy to GCS when I tried to use a GCS bucket as a remote cache. It needs to be updated.\r\n", "See https://github.com/bazelbuild/bazel/pull/8017" ]

[]

[ "P2", "type: process", "team-Remote-Exec" ]

[ "third_party/BUILD" ]

[ "third_party/BUILD" ]

[]

[ "timestamp(timedelta=36669.0, similarity=0.8402573714705566)" ]

[ "Hmm, I thought I had updated it: 9259683c6b460629d1ca83759ce26fa76839558d\r\n\r\nAre you building with any `--incompatible_` flags?", "I think the problem is that there's also `third_party/java/java_tools/turbine_deploy.jar`.", "I hacked up an integration test to demonstrate the problem:\r\n```\r\n$ git diff\r\ndiff --git a/src/test/shell/integration/java_integration_test.sh b/src/test/shell/integration/java_integration_test.sh\r\nindex 9435e13101..fc0790afea 100755\r\n--- a/src/test/shell/integration/java_integration_test.sh\r\n+++ b/src/test/shell/integration/java_integration_test.sh\r\n@@ -634,6 +634,7 @@ java_library(name = 'processor_dep',\r\n \r\n java_plugin(name = 'processor',\r\n processor_class = 'test.processor.Processor',\r\n+ generates_api = True,\r\n deps = [ ':annotation', ':processor_dep' ],\r\n srcs = [ 'Processor.java' ])\r\n EOF\r\n@@ -695,6 +696,10 @@ java_library(name = 'client',\r\n srcs = [ 'ProcessorClient.java' ],\r\n deps = [ '//$pkg/java/test/processor:annotation' ],\r\n plugins = [ '//$pkg/java/test/processor:processor' ])\r\n+\r\n+java_binary(name = 'bin',\r\n+ srcs = ['MyBinary.java'],\r\n+ deps = [':client'])\r\n EOF\r\n \r\n cat >$pkg/java/test/client/ProcessorClient.java <<EOF\r\n@@ -704,11 +709,19 @@ import test.processor.TestAnnotation;\r\n class ProcessorClient { }\r\n EOF\r\n \r\n- bazel build ${bazel_opts[@]+\"${bazel_opts[@]}\"} //$pkg/java/test/client:client --use_ijars || fail \"build failed\"\r\n+ cat > $pkg/java/test/client/MyBinary.java <<EOF\r\n+class MyBinary {\r\n+ public static void main(String[] args) {\r\n+ new ProcessorClient();\r\n+ }\r\n+}\r\n+EOF\r\n+\r\n+ bazel build ${bazel_opts[@]+\"${bazel_opts[@]}\"} //$pkg/java/test/client:bin || fail \"build failed\"\r\n unzip -l ${PRODUCT_NAME}-bin/$pkg/java/test/client/libclient.jar > $TEST_log\r\n expect_log \" test/Generated.class\" \"missing class file from annotation processing\"\r\n \r\n- bazel build ${bazel_opts[@]+\"${bazel_opts[@]}\"} //$pkg/java/test/client:libclient-src.jar --use_ijars \\\r\n+ bazel build ${bazel_opts[@]+\"${bazel_opts[@]}\"} //$pkg/java/test/client:libclient-src.jar \\\r\n || fail \"build failed\"\r\n unzip -l ${PRODUCT_NAME}-bin/$pkg/java/test/client/libclient-src.jar > $TEST_log\r\n expect_log \" test/Generated.java\" \"missing source file from annotation processing\"\r\n$ bazel test //src/test/shell/integration:java_integration_test --test_filter test_java_plugin\r\n```", "Thanks, I'd forgotten about #7196.\r\n\r\nIt's concerning that there apparently aren't any integration tests using `generates_api`.\r\n\r\n@lberki are there plans for more automation around `third_party/java/java_tools/update_java_tools.sh`?", "Closing since 92adff893fbd185269bf734abf4d3b12d3117ebc fixed the immediate problem. It does seem like there should be an integration test for javac turbine.", "I'm getting this error on 0.23.0 with `--all_incompatible_changes`." ]

[ "nit: Isn't this duplicated from the 0.22 (see line 38 in the same file?)", "The file didn't have a trailing newline, the only thing that changed on this line was trailing whitespace.", "Oh, also, I think the duplication with 38 is deliberate? The script is printing the list of files that change in each update, and this file changed in this update and the one below." ]

[ "team-Rules-Java" ]

[ "third_party/java/java_tools/README.md", "third_party/java/java_tools/update_java_tools.sh" ]

[ "third_party/java/java_tools/README.md", "third_party/java/java_tools/update_java_tools.sh" ]

[]

[ "timestamp(timedelta=0.0, similarity=0.8577877964005944)" ]

[ "Not team-Starlark?\r\nI'm tagging @c-parsons since `{s,}he` has opened the original deprecation issue. ", "It's probably my fault, sorry about that.\r\nI'm looking at it.", "@laurentlb I guess you did too much refactoring on https://github.com/bazelbuild/bazel/commit/2b3ad18eec4e6b968d840777249c809e478b2e13 and removed the warning with the flag by mistake. " ]

[]

[ "P1", "bad error messaging" ]

[ "src/main/java/com/google/devtools/build/lib/syntax/Identifier.java" ]

[ "src/main/java/com/google/devtools/build/lib/syntax/Identifier.java" ]

[ "src/test/java/com/google/devtools/build/lib/packages/PackageFactoryTest.java" ]

[ "timestamp(timedelta=0.0, similarity=0.9373832826710414)" ]

[]

[ "when does this happen?", "If `path` itself was empty.\r\nShall I move this check above the conversion?", "Yes please.", "Done." ]

[ "type: bug", "P1", "area-Windows", "team-OSS" ]

[ "src/main/cpp/util/file_windows.cc" ]

[ "src/main/cpp/util/file_windows.cc" ]

[]

[ "timestamp(timedelta=1.0, similarity=0.8840656578174207)" ]

[ "If I'm understanding my build errors correctly, this change re-broke the py2/3 compatibility on make_deb.py that #6818 had previously fixed.\r\n\r\n```\r\n File \".../bazel-out/k8-opt/bin/external/bazel_tools/tools/build_defs/pkg/make_deb.runfiles/bazel_tools/tools/build_defs/pkg/make_deb.py\", line 301, in GetFlagValue\r\n flagvalue = flagvalue.decode('utf-8')\r\nAttributeError: 'str' object has no attribute 'decode'\r\n```\r\n\r\nIt's possible that the toolchain changes in 0.25.0 are also at fault and/or the only problem, but with the binary tagged `PY2AND3`, the code doesn't look correct to me on first read.", "See https://github.com/bazelbuild/rules_pkg/issues/35" ]

[ "Can you add \r\n # -*- coding: utf-8 -*-\r\nat the top of this file and build_test.sh ", "I think it is safer to do the decode at the top of the method on the incoming value.\r\nOr, better yet, the transform from bytes to unicode string should happen in GetFlagValue.", "If you would like to do future maintainers a favor, you could clean up the formatting on the block above this.\r\nThe current formatting is an unreadable mess. How about.\r\n\r\n changesdata = ''.join(\r\n MakeDebianControlField(*x)\r\n for x in [\r\n ('Format', '1.8'),\r\n ('Date', time.ctime(timestamp)),\r\n ('Source', package),\r\n ('Binary', package),\r\n ('Architecture', architecture),\r\n ('Version', version),\r\n ('Distribution', distribution),\r\n ('Urgency', urgency),\r\n ('Maintainer', maintainer),\r\n ('Changed-By', maintainer),\r\n ('Description', '\\n%s - %s' % (package, short_description)),\r\n ('Changes', (\r\n '\\n%s (%s) %s; urgency=%s'\r\n '\\nChanges are tracked in revision control.') % (\r\n package, version, distribution, urgency)),\r\n ('Files', '\\n' + ' '.join(\r\n [checksums['md5'], debsize, section, priority, deb_basename])),\r\n ('Checksums-Sha1', '\\n' + ' '.join(\r\n [checksums['sha1'], debsize, deb_basename])),\r\n ('Checksums-Sha256', '\\n' + ' '.join(\r\n [checksums['sha256'], debsize, deb_basename]))])\r\n\r\nOr even:\r\n changesdata = ''.join(\r\n MakeDebianControlField('Format', '1.8'),\r\n MakeDebianControlField('Date', time.ctime(timestamp)),\r\n ...\r\n\r\nIf not, I'll clean it up after.", "The decode() could be at line 295 \r\n return flagvalue.decode('utf-8')\r\n\r\nBut also we might want to add it at line 289\r\n flagvalue = flagvalue.decode('utf-8')\r\nso that an @file usage can have a utf-8 path", "Thanks for the review aiuto! Apologies it's taken me so long to get back to you. I'd be happy to add the encoding.\r\n\r\nI was seeing `# -*- coding: utf-8 -*-` in other places in the codebase. (Mainly third_party/protobuf) Is that preferred or is `# -- coding: utf-8 --` alright?", "That's fair. I'm happy to do it. I was trying to keep my initial changes minimal so it was easier to review.\r\n\r\nI like the below approach so will go ahead and do that unless you prefer the other method:\r\n```python\r\nchangesdata = ''.join([\r\nMakeDebianControlField(...),\r\nMakeDebianControlField(...),\r\n...\r\n])\r\n```", "Sounds good. I'll play around with this a little. I had a friend suggest setting `changesdata = u''` and `changesfile = u''` as well so they won't have to be overridden at each access point.", "-*-. My typo the firs time.", "You might as well do the encoding just once. ", "That works for me too. It is easier to follow than \"MakeDebianControlField(*x)\"", "Doing some testing - it seems that doing:\r\n```python\r\nif flagvalue:\r\n flagvalue = flagvalue.decode('utf-8')\r\n ...\r\n```\r\nand removing the decode in: `return result.decode('utf-8').replace(...) + '\\n\\'` on line 145 will cause UnicodeDecodeErrors if one of the fields that doesn't use GetFlagValue(s) in CreateDeb (line 306) contains a utf-8 variable.\r\n\r\nI could move the `flagvalue = flagvalue.decode('utf-8')` line further into the if statement nest so the `@file` directive will get it and leave the rest the same.\r\n", "I agree. However, I get a TypeError from the `f.addfile(... BytesIO(controlfile))` line if I do the encoding at the top of the method. It will work if I do:\r\n```python\r\ncontrolfile = ''\r\n for values in DEBIAN_FIELDS:\r\n fieldname = values[0]\r\n key = fieldname[0].lower() + fieldname[1:].replace('-', '')\r\n if values[1] or (key in kwargs and kwargs[key]):\r\n controlfile += MakeDebianControlField(fieldname, kwargs[key], values[2])\r\n controlfile = controlfile.encode('utf-8')\r\n...\r\n```\r\nand remove the encode() calls further down.\r\n\r\nDo you have any idea why that would be? My guess would be the `controlfile += MakeDebianControlField(...)` line in the `for` loop but I'm unsure why that would override the encoding if it's set earlier. It seems like both this TypeError and the issues I run into `MakeDebianControlField` function result from the `installed_size` flag.", "I think that is worth it. We have other outstanding bugs for not supporting UTF-8 file paths. So GetFlagValue needs to do that. AND the content should be interpreted as utf-8 encoded.\r\n\r\nBut I think I have the solution to what you observed. We upgrade str to unicode str in MakeDebianControlField\r\nI changed one of the non GetFlagValue() attributes to be unicode also, so the test would have a that case too.\r\n maintainer = \"somé[email protected]\",\r\n\r\nIf you have extra tests, I would like to try them against my suggestion. If you think that change works. I also have it patched on top of your PR, so I could to the merge of yours.\r\n\r\nFull diff: https://github.com/bazelbuild/bazel/compare/master...aiuto:deb\r\n\r\ninline diff\r\n\r\n*** 133,148 ****\r\n def MakeDebianControlField(name, value, wrap=False):\r\n \"\"\"Add a field to a debian control file.\"\"\"\r\n result = name + ': '\r\n if type(value) is list:\r\n! value = ', '.join(value)\r\n if wrap:\r\n! result += ' '.join(value.split('\\n'))\r\n result = textwrap.fill(result,\r\n break_on_hyphens=False,\r\n break_long_words=False)\r\n else:\r\n result += value\r\n! return result.decode('utf-8').replace('\\n', '\\n ') + '\\n'\r\n \r\n \r\n def CreateDebControl(extrafiles=None, **kwargs):\r\n--- 133,150 ----\r\n def MakeDebianControlField(name, value, wrap=False):\r\n \"\"\"Add a field to a debian control file.\"\"\"\r\n result = name + ': '\r\n+ if type(value) is str:\r\n+ value = value.decode('utf-8')\r\n if type(value) is list:\r\n! value = u', '.join(value)\r\n if wrap:\r\n! result += u' '.join(value.split('\\n'))\r\n result = textwrap.fill(result,\r\n break_on_hyphens=False,\r\n break_long_words=False)\r\n else:\r\n result += value\r\n! return result.replace('\\n', '\\n ') + '\\n'\r\n \r\n \r\n def CreateDebControl(extrafiles=None, **kwargs):\r\n***************\r\n*** 295,303 ****\r\n \r\n def GetFlagValue(flagvalue, strip=True):\r\n if flagvalue:\r\n if flagvalue[0] == '@':\r\n with open(flagvalue[1:], 'r') as f:\r\n! flagvalue = f.read()\r\n if strip:\r\n return flagvalue.strip()\r\n return flagvalue\r\n--- 297,306 ----\r\n \r\n def GetFlagValue(flagvalue, strip=True):\r\n if flagvalue:\r\n+ flagvalue = flagvalue.decode('utf-8')\r\n if flagvalue[0] == '@':\r\n with open(flagvalue[1:], 'r') as f:\r\n! flagvalue = f.read().decode('utf-8')\r\n if strip:\r\n return flagvalue.strip()\r\n return flagvalue\r\n\r\n\r\n\r\n", "Sounds good - if there are other outstanding utf-8 filepath bugs it'd be worth it to deal with that while we're here. \r\n\r\nAmusingly - that was the exact test case I was using. I'd be fine with you merging on top of my commits. I guess I didn't go far enough in changing things to unicode strings!\r\n\r\nThe one other thing I was noticing was that if the controlfile isn't utf-8 encoded when determining the size content will be dropped from it. I've been using `expect_log \"Built-Using: Bazel\"` to test for that as that seems to usually be the last line in the controlfile. However, I'm not sure how robust of a test that is.", "Yes. I saw your fix for that. I'll try to commit this now. Having flakiness with the windows/jdk8 tests. So this may take a few tries to catch a clean build.", "Looks like it passed. Thanks for the review!" ]

[ "P4", "team-Rules-Server" ]

[ "tools/build_defs/pkg/BUILD", "tools/build_defs/pkg/build_test.sh", "tools/build_defs/pkg/make_deb.py" ]

[ "tools/build_defs/pkg/BUILD", "tools/build_defs/pkg/build_test.sh", "tools/build_defs/pkg/make_deb.py" ]

[]

[ "timestamp(timedelta=160224.0, similarity=0.8566001287297237)" ]

[ "(Plan is to use this work to also address #4815, rather than implement an ad hoc solution to that bug.)", "Design doc [here](https://github.com/bazelbuild/rules_python/blob/master/proposals/2019-02-12-design-for-a-python-toolchain.md).\r\n\r\nDiscussion can proceed in this thread. Initial reviewers are @katre (Bazel, Toolchains), @mrovner (Python), and @nlopezgi (Remote Execution).", "Overall, this looks good. A few points:\r\n\r\n1. Get rid of the host_python_toolchain, just use the autodetecting version. I don't think the time spent running the autodetection script will really hurt python performance, will it?\r\n1. In the **Backwards Compatbility** section, you mention deleting several flags: I do not think flags can be conditionally deleted based on another flag. Ignoring the values should be sufficient, and they can be deleted in a post-migration cleanup.", "1. Ok, I'll remove `host_python_toolchain`. I don't have a strong opinion about its impact on performance, but it's worth remembering that it will still be superseded by any other applicable toolchain that the user registers.\r\n\r\n2. By \"delete\" I really mean \"fail-fast if the user tries to specify them while the incompatible flag is enabled\". This can be implemented by checking the flag value for null, and logging an error in the fragment's `reportInvalidOptions()` (see [example](https://github.com/bazelbuild/bazel/blob/ea956ae20bac5be1e9917faabdcac3cab2965c12/src/main/java/com/google/devtools/build/lib/rules/python/PythonConfiguration.java#L99)). The true deletion would happen around the time the old code paths and incompatible flag are removed.", "I think the distinction between \"delete\" and \"ignore with a warning\" is worth calling out.", "(See also Nicolas's comments [here](https://github.com/bazelbuild/rules_python/pull/161#pullrequestreview-202814381).)", "Replying to @nlopezgi's concern about toolchains and extensibility here:\r\n\r\n> I think you need something similar as there are rules that execute the native py_binary (e.g., rules_docker does here: https://github.com/bazelbuild/rules_docker/blob/master/python/image.bzl#L84) and there might be something special that these type of users need to do to work with the new py toolchains properly.\r\n\r\nUsers who operate at the level of BUILD targets and macros only need to worry about having definitions for their platforms and python toolchains, i.e. the `py_runtime_pair` target and an associated `toolchain` target referencing it, with appropriate constraints. Whatever the build graph looks like, it should Just Work.\r\n\r\nUsers who are writing new rules that interoperate with the Python rules may need to manipulate Python toolchains. To receive a Python toolchain, they should declare `@bazel_tools//tools/python:toolchain_type` as a required toolchain type in their rule and receive it with `ctx.toolchains[<same label>]`. This would give them back the `ToolchainInfo` with the schema described in the doc. This would be necessary for instance to write a replica of `py_binary` in Starlark. Conversely, users can write a replacement for `py_runtime` or `py_runtime_pair` by simply returning an appropriate `PyRuntimeInfo` / `ToolchainInfo` provider.\r\n\r\nI didn't go into too much detail about how to write rules that produce or consume Python toolchains (aside from a description of the provider schema) because this is already covered by the general toolchain documentation linked in the doc's abstract.", "(please do not assign issues to more than one `team-*`)", "Let me add my 5 coins as an end user\r\nI would like to see similar naming convention and signatures for toolchain types as is done for all other toolchains\r\n```\r\npy_runtime_pair -> py_toolchain\r\n```\r\nAlso related the section\r\n> Why not split Python 2 and Python 3 into two separate toolchain types?\r\n> The naming convention for toolchain types is to literally name the target \"toolchain_type\",\r\n\r\nBased on jdk toolchain implementation it doesn't look like a problem - [tools/jdk/BUILD](https://github.com/bazelbuild/bazel/blob/be9483129401a7c9f1a86a40a3d7770662248576/tools/jdk/BUILD#L25)\r\nAlso \r\n```\r\n@bazel_tools//tools/python2:toolchain_type\r\n@bazel_tools//tools/python3:toolchain_type\r\n```\r\ndoesn't break the convention but helps avoids problems with a missed py_runtime attribute in the pair and provide an opportunity to force version as \r\n```\r\n--python3_toolchain .../python:remote3.7.15 --python2_toolchain ../python:remote2.7.13\r\n```\r\nAlso it allows to rid of python_version attribute from `py_binary`/`py_test` rules and select runtime based on provided toolchain type\r\n", "Hi Guro, thanks for the comments. There's a few reasons I didn't organize it that way.\r\n\r\n> py_runtime_pair -> py_toolchain\r\n\r\nIt's not clear what the ideal way to deal with multiple Python runtimes will be long-term. Currently the rules are organized around this PY2/PY3 distinction, but in the future that will become less important and there may be more emphasis on choosing minor versions. Building the current schema into the rule name gives us a little more room to redesign it in the future without extra breaking changes.\r\n\r\n>> Why not split Python 2 and Python 3 into two separate toolchain types?\r\n>\r\n> Based on jdk toolchain implementation it doesn't look like a problem\r\n\r\nAgain, future-proofing is part of the reason. It's easier to migrate the schema of a `ToolchainInfo` to use some other layout rather than `py2_runtime = ..., py3_runtime = ...`, and even to migrate a toolchain rule to have different attributes, than it is to migrate Python rules to accept different toolchain types.\r\n \r\nFor platforms that do not provide an implementation for one of these toolchains, we'd still have to define a dummy toolchain that fails at analysis time, since even if a target does not use the toolchain its rule type will still require both toolchain types.\r\n\r\nAlso, creating `.../python[2|3]:toolchain_type` implies the creation of two additional directories and `BUILD` files whose only purpose is to provide naming disambiguation for the toolchain types. I'd sooner just break convention and name them `.../python:py[2|3]_toolchain_type`.\r\n\r\n> provide an opportunity to force version\r\n\r\nI don't believe there's anything in the toolchain framework that would allow you to directly force the toolchain implementation to use for a given toolchain type, as opposed to just letting Bazel select one based on constraints, though I could be wrong.\r\n\r\n> Also it allows to rid of `python_version` attribute from `py_binary`/`py_test` rules and select runtime based on provided toolchain type\r\n\r\nI checked with @katre before starting the design doc, and it turns out that currently toolchains can really only differentiate themselves based on platform constraints, not other rule logic constraints. You can't easily express that a given py_binary has a constraint of \"I want Python 2\".\r\n", "Looks good to me.", "Thanks for addressing my question about extending python rules. It would be ideal to get this info as part of user docs whenever the right time for that comes.\r\n\r\nLooks good to me. ", "@brandjon is it really not possible to express a version constraint for e.g. a py_binary using toolchains? There is an example here: https://docs.bazel.build/versions/master/platforms.html#defining-constraints-and-platforms talking about defining Version as a constraint. That could not be used?\r\n\r\nThis is sidetracking a bit but I am also curios about this as I was planning to implement to add a constraint for the nodejs toolchains so it would be possible to use different versions of nodejs in the same workspace and one could choose on nodejs_binary level. ", "In that example the platform either supports one version or the other, and targets build/run on whichever platform has the version they require. Here the same platform is generally expected to provide both Python versions. Currently all values for a constraint setting are considered mutually exclusive.\r\n\r\nI think the bigger issue is that Python version is a property of the target, not the platform. @katre could comment more.", "Ah yeah I see, so I can say a platform is compatible with a certain version, but I can not say that a platform is compatible with multiple versions and then just choose the version on a per-target level. It would be nice to hear more details about how to make such a use-case possible in a generic way from @katre. As I imagine this could come up again and again. I am also happy to pull this in a separate issue or bazel-discuss as this is just tangentially related to this issue.\r\n\r\nI also do wonder how `rules_go` manages to allow the setting toolchain attributes on a per-target basis, e.g. I can set `go_binary(os=\"windows\",...)` and then that will compile for windows.", "We don't yet have a good story for handling constraints on specific toolchain versions. I know that rules_scala also has problems with this, but I am not sure what the best UX would be (ie, what is the easiest way for users to express the idea that \"this target needs nodejs X.Y to build\").\r\n\r\nI don't think there's an existing issue in Github to track this, so feel free to open one and make any suggestions as to how you think it should be handled.", "I'm running into some bootstrapping issues with the new provider definition that I won't have time to address before we cut the baseline for 0.24. So this feature and the associated --incompatible change to preview it will be available in 0.25.", "Legacy code cleanup tracked by #8169.", "Reopening this because the flag flip failed due to downstream breakages. Follow the migration tracking bug #7899.", "This seems to be resolved, but as a casual user of Python and new user of rules_python, it would be helpful to have simple documentation explaining how to do a hermetic build.\r\n\r\n(Somehow I am able to run my py_binary targets at the moment even though I know the dependencies in the BUILD.bazel files are inadequate.)" ]

[ "Can the input `bash_binary` ever be just `bash.exe` for example, which would mean \"whatever is on the PATH\"?", "I think there are two ways to tell which shell binary to use, `BAZEL_SH` or `--shell_executable`. Both way requires user to set an absolute path. The only case `bash_binary` is a relative path (runfile path) is that when using shell toolchain (similar to python toolchain), but I'm not sure it's implemented.", "But here, `python_binary` will just be `python` if nothing is specified.. I need to fix this.", "Thanks, let me know when the PR is ready for another review.", "Thanks for catching the problem! Please take a look again. ;)", "Just checking: `bash_binary` can either be an absolute Bash path (from `BAZEL_SH` or `--shell_executable`), or (theoretically perhaps) a bash toolchain's runfiles-relative path. Correct?\r\n\r\nAnd it always ends with \"bash\" or \"bash.exe\" (so just \"bash\" without the extension), right?\r\n\r\nAny other option I'm missing?", "What other values could `python_binary` have, and in what situations?", "Everything you said I believe is true, I don't think there's any other option.", "Similar to shell, we have\r\n\r\n1. `--python_path`, gives an absolute path\r\n2. `--python_top`, gives an runfile path\r\n3. python toolchain (`py_runtime`), gives a runfile path\r\n4. Default if nothing specified: `python`\r\n\r\n@brandjon Can confirm this.", "Thanks for confirming that!", "Could you please add a comment about this?", "Thanks! Waiting for @brandjon to confirm before LGTM. Please add a comment in the code summarizing Jon's answer.", "Will do that!", "Oh, just realized I had a typo. It should be `GetBinaryPathWithoutExtension(bash_binary) != L\"bash\")`, `GetBinaryPathWithoutExtension` just strips out the ending `.exe` if there is any. I wanted to check if bash_binary is already `bash` or `bash.exe`. Similar for the python case.", "Will add comment for this!", "Both `--python_top` and Python toolchains select a `py_runtime` to use. `py_runtime` provides a runfiles path if its `interpreter` attribute is set, and an absolute path if its `interpreter_path` attribute is set (these attributes are exclusive).\r\n\r\n`--python_path` is validated to be an absolute path. If nothing is specified, the `python` default is passed along the same code path that `--python_path` uses but skips this validation check as a special case. It is intended that both of these approaches are deprecated and replaced by toolchains, with the `python` PATH lookup behavior provided by an autodetecting toolchain.", "Thanks for the confirmation, I'll add those to comment", "Rewrite comment as follows:\r\n\r\n> There are three kinds of values for `python_binary`:\r\n> 1. An absolute path to a system interpreter. This is the case if `--python_path` is set by the user, or if a `py_runtime` is used that has `interpreter_path` set.\r\n> 2. A runfile path to an in-workspace interpreter. This is the case if a `py_runtime` is used that has `interpreter` set.\r\n> 3. The special constant, \"python\". This is the default case if neither of the above apply.\r\n> Rlocation resolves runfiles paths to absolute paths, and if given an absolute path it leaves it alone, so it's suitable for cases 1 and 2.\r\n\r\nThere's a slight detail that the default toolchain's py_runtime on windows contains a sentinel value hack that tells bazel to use legacy --python_path, but I don't think we need to mention that in this comment.", "Thanks for help with the comment!" ]

[ "type: feature request", "P1", "team-Rules-Python" ]

[ "src/tools/launcher/bash_launcher.cc", "src/tools/launcher/python_launcher.cc" ]

[ "src/tools/launcher/bash_launcher.cc", "src/tools/launcher/python_launcher.cc" ]

[]