[6.3.0] Adjust --top_level_targets_for_symlinks

[iancha1992]

Adjust --top_level_targets_for_symlinks.

• If all targets in $ bazel build //... have the top-level config, bazel-bin, etc. point to that config

• If all targets in $ bazel build //... have the same transitioned config, bazel-bin, etc. point to that config

• If targets in $ bazel build //... have mixed configs, bazel-bin, etc. are deleted

• If all targets in $ bazel build //... have the top-level config, bazel-bin, etc. point to that config

• If all targets in $ bazel build //... have the same transitioned config, bazel-bin, etc. point to that config

• If targets in $ bazel build //... have mixed configs and at least one of them is the top-level config, bazel-bin, etc. point to the top-level config

• If targets in $ bazel build //... have mixed configs and none are the top-level config, bazel-bin, etc. are deleted

Fixes #17081.

Commit ceb9955

Closes #18854.

PiperOrigin-RevId: 546938509
Change-Id: I75adf0b8c2094522125c5e65d8c450eb2436d392

[iancha1992]

@fmeum @gregestren Looks like the tests are failing. Do you know why?

cc: @bazelbuild/triage

[gregestren]

@fmeum @gregestren Looks like the tests are failing. Do you know why?

cc: @bazelbuild/triage

1cd3588 changed the affected code's method signature. I'd tentatively suggest merging that first but it looks like @joeleba has a bunch of commits on that file: https://github.com/bazelbuild/bazel/commits/26926ee068ce7c18ec9a8afee49e078fa7ab6a1c/src/main/java/com/google/devtools/build/lib/buildtool/ExecutionTool.java I'd check with him to see what's safe.

If necessary we can probably merge this independently but we'd have to refactor some amount of code.

[gregestren]

If necessary we can probably merge this independently but we'd have to refactor some amount of code.

I really just glanced at this but I suspect a tiny adjustment to this PR would also suffice, without having to worry about all those other commits. I'd still like to get @joeleba 's input first.

[joeleba]

Sorry, I'm not familiar with this process. Are we trying to cherrypick your commit ceb9955 on top of a baseline foo, where foo comes before my commit 1cd3588?

@joeleba has a bunch of commits on that file: https://github.com/bazelbuild/bazel/commits/26926ee068ce7c18ec9a8afee49e078fa7ab6a1c/src/main/java/com/google/devtools/build/lib/buildtool/ExecutionTool.java I'd check with him to see what's safe.

It should be safe to merge 1cd3588. The other commits in the chain are independent of this.

[gregestren]

Sounds good. @iancha1992 can we merge 1cd3588 first? Then this one should resolve.

[gregestren]

Sorry, I'm not familiar with this process. Are we trying to cherrypick your commit ceb9955 on top of a baseline foo, where foo comes before my commit 1cd3588?

Yes.

[iancha1992]

Sounds good. @iancha1992 can we merge 1cd3588 first? Then this one should resolve.

Looks like there's a conflicts with

1. src/main/java/com/google/devtools/build/lib/buildtool/BuildTool.java
   function "buildTargetsWithMergedAnalysisExecution", currently has the try block, and is quite different from the master's. For example, "() -> executionTool.prepareForExecution(request.getId()));", instead of () -> executionTool.prepareForExecution(request.getId(), executionTimer). And also "if (analysisAndExecutionResult == null)" does not exist currently in the same try block form the file. We may need another commit that address this change from the "buildTargetsWithMergedAnalysisExecution's try block".

@gregestren can you please take a look?

[gregestren]

I'm going to manually resolve and move this over to #18916.

[iancha1992]

Closing this now because new cherry-pick now pass the tests in: #18916