ICP is one of the biggest Rust open source repositories using bazel.

Closes #24025.

PiperOrigin-RevId: 687427480
Change-Id: Ie340064b714987f4d9aace58dbd5a1f6876d492b

The toolchain context of the aspect is currently used to configure the dependencies of the base target, which is incorrect and results into a number of issues due to missing execution groups or different ones being picked up.

This CL modifies `AspectFunction` to always prepare the `baseTargetUnloadedToolchainContexts` and pass it to `PrerequisiteParameters`. `PrerequisiteParameters#baseTargetUnloadedToolchainContexts` is then used for the target's dependencies while `PrerequisiteParameters#toolchainContexts` continues to be used for the aspect owned dependencies.

The base target toolchains and execution groups will not be available in the aspect implementation function via `ctx.rule.toolchains` and `ctx.rule.exec_groups` unless the aspects path propagates to toolchains.

PiperOrigin-RevId: 687376763
Change-Id: I78167bab7f8648e5855178d5dbcc3fcb02357a38

PiperOrigin-RevId: 687311869
Change-Id: I2f481f7a702494691093bf91905fbb28f309a482

Removing this assumption makes it possible to put something into the default MODULE.bazel, generated by setup_module_dot_bazel.

In my case, I'll use this to temporarily patch rules_java and protobuf so that they don't cause additional unnecessary downloads.

Some tests assumed that MODULE.bazel was empty and deleted its content. Other tests needed empty MODULE.bazel, and assumed it already was.

I believe this change is generally useful, when we need to do some patching.

PiperOrigin-RevId: 687256923
Change-Id: Id0a5f25676c0ad8992737a37dfe9acaf573f1d35

Fixes https://github.com/bazelbuild/bazel/issues/16496

Closes #24019.

RELNOTES: The stripPrefix parameter of repository_ctx.download_and_extract()
and repository_ctx.extract() has been renamed to strip_prefix; the deprecated
stripPrefix name remains usable for compatibility.
PiperOrigin-RevId: 687224164
Change-Id: Iffaba2e65c049a2ff8bb98d1fba52e0bba9e5a02

PiperOrigin-RevId: 687111838
Change-Id: Ifb804d115aa0adec833287e67f44c519080c299b

Remove word to fix grammar issue.

Closes #24018.

PiperOrigin-RevId: 687041869
Change-Id: I2802ea9f6ddb1503f48e9c1f5f78d2446954ca3f

When switching to cpp linking variables, we need to modify the Apple rules to
ensure that -Wl,-rpath,/usr/lib/swift appear before other rpaths.

But we still need to leave behind that rpath in swift rules for linking rules
that don't use apple rules (e.g. cc_binary).

The side effect is that the are now more Appler linker warning about duplicate
rpaths for simple targets.  I don't know of a way to turn that off, and since we
already have logic to deduplicate frameworks, let's deduplicate rpaths while
we're at it.

PiperOrigin-RevId: 687016661
Change-Id: Iba80b5690121ba90453e461057167d0920eafa5f

The external repository test depends on a .tgz file which would be downloaded from another repo. I think that when whomever made the original test, they just copied six-1.0.0.tar.gz to the test data. But the test does not actually care about what the repo does. It only cares that it can unload it and inject a BUILD file.  This PR replaces that tarball with a trivial equivalent.

This does not impact the bazel distribution size, it just means we have one less false dependency.

Fixes: #18272

Closes #23875.

PiperOrigin-RevId: 686978582
Change-Id: Id59be40caa6625e6bd39d36cef108ee15b3a8d9e

Previously we weren't creating input files for labels used in the attribute of a top-level symbolic macro. This meant that if you wrote

```
my_macro(
    name = "foo",
    srcs = ["input.txt"],
)
```

you'd get an error message claiming that input.txt is not declared and requires an `exports_files()`. (The usages of input.txt by targets declared inside of foo don't count, because implicit input file creation only works for usages of labels outside of symbolic macro bodies.)

This CL fixes this behavior, refactors the relevant logic in Package.java, and adds more test coverage.

Package.java:
- Factor the big for loop from `beforeBuild()` to `createAssumedInputFiles()`. Leave the original for loop in place for test suite accumulation, and pull the `if (discover...)` out of the new call.
- Factor the meat of the logic into a helper, `maybeCreateAssumedInputFile()`, so we can reuse it for both rules and macros (previously there was no loop over references in macros).
- Add tedious javadoc to this tedious logic.

SymbolicMacroTest.java
- Add explanatory comment to `macroCanReferToInputFile()`.
- Expand `macroCannotForceCreationOfImplicitInputFileOnItsOwn()` to check that the target is still not created when the inner usage is in an attr of a MacroInstance rather than a Rule.

PackageFactoryTest.java
- Expand comment in `testSymbolicMacro_macroPreventsImplicitCreationOfInputFilesUnderItsNamespace()`, include coverage for when the reference comes from within a target or submacro inside the symbolic macro itself.
- New test case for the behavior of this CL, `..._macroInstantiationCanForceImplicitCreationOfInputFile()`.
- Drive-by: Add a test case for when a badly named target in a symbolic macro clashes with an implicitly created input file. This behavior may be affected by lazy macro evaluation in the future.

Confirmed that this CL works for the case I originally discovered it on while writing examples documentation.

Fixes #24007.

PiperOrigin-RevId: 686919555
Change-Id: I6313070f76456c0b0b5d5458ca35c89d1d6da33b

The watchdog is setup at the beginning of the analysis/execution phase, by listening to an event. It's then cleaned up at the end of the build, via a `finally` clause.

If the build fails early enough in the analysis/execution phase, the cleanup code would win the race and be triggered before the watchdog was setup.

PiperOrigin-RevId: 686852796
Change-Id: I736a17312c601c3720e2ac751d87357c3faec08f

Finishing touches for #4867.

RELNOTES: None.
PiperOrigin-RevId: 686819148
Change-Id: I9ff2424683b8579ad5a953f7974c9579c5e715db

This change adds a "--quiet" startup option, which causes Blaze not to print any status messages, progress messages, info messages or warnings of any kind. Error messages are still emitted as usual.

Fixes #4867.

RELNOTES[NEW]: The "blaze --quiet" command line option can now be used to make Blaze emit much less output.

PiperOrigin-RevId: 686784300
Change-Id: Ibaa0b6a1788b43863337dce6fc16da09defb6724

PiperOrigin-RevId: 686659518
Change-Id: I4a9de1ac4880ef1c5dd01e4b5bcb1538e69029af

Fixes #21884
Fixes #23053

Closes #23215.

PiperOrigin-RevId: 686518093
Change-Id: I1a6699b169fe4291ad0dda3be7e74f38c8e37bf1

PiperOrigin-RevId: 686495724
Change-Id: I07c07ec72cd461e74a0ec959d074abf4aaeaf7e6

Fixes #23981

Closes #23982.

PiperOrigin-RevId: 686486627
Change-Id: I5f03b63601a535f8d551c005b00bf040846a324d

This really just exposes the existing LABEL_DICT_UNARY type that already exists in Java.

Other than writing some boring conversion logic, all pieces fell together pretty pleasantly.

Work towards #7989. That one calls for a string-to-label-list-dict type, but I'm planning to resolve that as WAI, since string-to-label-dict is close enough.

RELNOTES: None.
PiperOrigin-RevId: 686415025
Change-Id: Ib07ada7ab2ede95220ed1cc2d3569996fc4afb88

PiperOrigin-RevId: 686412821
Change-Id: I7c8410885474ef8aec6603468f32283178625f83

This required a number of somewhat awkward .toRelative() calls, but the alternative would have been to make IgnoredSubdirectories either have all absolute paths or all relative paths (both prefixes and inputs) and this one seemed simpler.

Progress towards #7093.

RELNOTES: None.
PiperOrigin-RevId: 686411349
Change-Id: I62b0b7f12801e7b44b6e064893c1dc0cb282badb

In case of targets not using toolchain resolution, the evaluated `ToolchainCollection<UnloadedToolchainContext>` is set to null but it can also be null if its value is not evaluated and needs a Skyframe restart. This CL returns a boolean from `getBaseTargetUnloadedToolchainContexts` to make it clear when a Skyframe restart is needed.

PiperOrigin-RevId: 686261289
Change-Id: Icb39aa89e212c39adff311341bc739e1f4b45223

When --enforce_project_configs is enabled for a valid --scl_config, do not allow any other flags to be set in the user blazerc (~/.blazerc or --blazerc) or command line.

PiperOrigin-RevId: 686252332
Change-Id: I20a1de3a261b9b75141ebe1661ee08e6e6e732d3

This PR address the #23932 issue with remote downloader.

Closes #23970.

PiperOrigin-RevId: 686180819
Change-Id: Ia36c5793622bd1ac1fdaa9ef1326ddc385a5a01f

PiperOrigin-RevId: 686177057
Change-Id: Id998df883ce6d8c977799194b6a9a3ad23f406fb

When cancelling an asynchronous repository download task, an interrupt signal is sent to the download thread. This doesn't mean that the download stops immediately. Avoid restarting a download until the previous download has actually stopped, so that the new download is able to clean old data without crashing (on Windows).

Fixes #21773

Closes #23837.

PiperOrigin-RevId: 686175953
Change-Id: I8d75f905b739d38b6cb430d5b5e84fda9a2d14e3

This is a more accurate name for the method, as it indicates that the node's version changed at the current version, regardless of whether its value changed.

PiperOrigin-RevId: 686173970
Change-Id: I6502aa611e286dcca3bccf3cbb7712e7794cecc4

PiperOrigin-RevId: 686133746
Change-Id: I98c6f89a7e0ab09e1fb19919ff9ef3af6b9ba204

Expose internal functions needed by multi_arch_binary_support.bzl and linking_support.bzl (both from builtins_bzl/common/objc/) via apple_common.

This is inpreparation for moving the content of both bzl files to the repo in unknown commit.

PiperOrigin-RevId: 686100272
Change-Id: Iadbe8da6a4335e30daca889cd19afcb627df2b5c

Progress towards #7093.

RELNOTES: None.
PiperOrigin-RevId: 686088272
Change-Id: I17ea2481ed58b76352ff1939b78ba93b8bdc9eac

Those are needed to remove proto rules from Bazel. Older version uses native.cc_proto_library in defs.bzl. Newer uses the rules from protobuf.

Additional dep on cc_proto_library_bzl is needed because rules_python use @rules_cc//cc:defs.bzl, which depends on cc_proto_library_bzl.

PiperOrigin-RevId: 686077918
Change-Id: Ie8ddfffd106a3e2be1a73a03dd27110064b6d917

Progress towards #9037.

RELNOTES: None.
PiperOrigin-RevId: 686074081
Change-Id: I919a452eb27d149b8e52c9c8188df294521fe347

This is needed to remove proto rules from Bazel. The automatic loading is disabled in unit tests.

PiperOrigin-RevId: 686069253
Change-Id: I8a2e5751ab98f24be1dc5e7854a57f68a06a6323

It's simpler.
Remove starlarkCcProtoAspect. It became unused.

PiperOrigin-RevId: 686068336
Change-Id: Id796a274c937ec1d433a7e9d8952783e4ec8f67e

Document the valid options for `--experimental_build_event_upload_strategy` for Bazel users in the option's help string.

In Bazel, this option may be either 'local' or 'remote'.

Fixes #23789.

RELNOTES: None.
PiperOrigin-RevId: 686067559
Change-Id: Ib7fde8c1210dea25cce30eaf1613cf0130ba037c

Progress towards #9037.

RELNOTES: None.
PiperOrigin-RevId: 686063047
Change-Id: If27b64213ed51f0b5f8c361c657b9bba2d7511f6

Setup copying. This will make it possible to use rules implementation from @protobuf repository in the tests.
Remove mocking of proto_lang_toolchain. It's not needed anymore
Rename protobuf_workspace to third_party/protobuf. This make loads nicer.
Keep mocking the rules. Tests are still using built-in rules and mixing them with @protobuf rules doesn't work well.
Remove some other unnecessary mocks.

Fix bazel_skylib to be imported in one go, instead of from the lib and rules directory.

PiperOrigin-RevId: 686052269
Change-Id: I9e7bcbcd1ce42fd27e88e4e446e74e9aec1605f8

This is the mechanical part of the change. It's not, strictly speaking, necessary, but given that the that set of path fragments will not be interpreted as a set of path fragments soon, I think it's better to be explicit.

An ancillary benefit is that now one can search for the new object to figure out which parts of the code handle the concept of "ignored subdirectories".

Progress towards #7093 .

RELNOTES: None.
PiperOrigin-RevId: 686046437
Change-Id: I4191297c9b1e5a6599d4c5f8eb2c80fae14ac220

Before this change, returning non-Label's from a computed default for `label_list` attributes would crash Bazel (at a later point while visiting the labels).

PiperOrigin-RevId: 685990009
Change-Id: I6b9c220a92f355d61c7340a6c383edc007f77be2

Record's implementation of hashCode and equals is inappropriate here due
to the recursive nature of these types. Instead, use reference equality.
These types are effectively interned via FileDependencyDeserializer
anyway.

PiperOrigin-RevId: 685830037
Change-Id: I00a28e80fa61805edd841f15391ea123c70af71e

PiperOrigin-RevId: 685784112
Change-Id: I7472b8789b5d14e8f0b898ab716eb4a22103d15f