cherry-pick of commit https://github.com/bazelbuild/bazel/commit/c04a5617d2c960e19177d6299f00a81624857925

The remarks about files being watched only if they're labels are no longer accurate.

Repo rule restarts are gone by default (`--experimental_worker_for_repo_fetching` defaults to `auto`), so the section doesn't really apply anymore.

Fixes https://github.com/bazelbuild/bazel/issues/22494

PiperOrigin-RevId: 668317824
Change-Id: Iba380b944cce2b5c8e7bd07606b89035ab401e72

Fixes:
```
syntax error at '"sha256"': expected }
```

Closes #23027.

PiperOrigin-RevId: 653606521
Change-Id: If52071612ab09bf5c0d77fbde228f9e664569d75

`http_archive` is not local - it does not re-download external repos on bazel server restart. Having `local=True` in the example is somewhat confusing.

Looks it was leftover after some older change:
https://github.com/bazelbuild/bazel/commit/13ecdf583301a94484a1ae0eb27c56fcf3248dc5#diff-1f760cde585b814b5608e64ade35cfa8f6fc0b1082316adc2f1bac412169a10eL41

Closes #22084.

PiperOrigin-RevId: 627954043
Change-Id: Ie25f5f0141401e194be99c9d02ec81c0ab619484

- Most references to WORKSPACE are changed to MODULE.bazel & friends
- Added information about the order in which Bazel looks for toolchains in MODULE.bazel + WORKSPACE
- More consistent terminology (eg. 'workspace' -> 'repo' when appropriate)
- More info about how `mirrors` in `bazel_registry.json` works

PiperOrigin-RevId: 595179646
Change-Id: I37aff58a2723aa830efeafb589fcf770e07b3711

When linking a nodeps dynamic library that contains object files
produced from tree artifact sources, the library may not contain any
references to the symbols in those files. In this case, when wrapping
the files in `--start-lib/--end-lib`, they may end up being dropped by
the linker, which results in missing symbols compared to a build with
`--dynamic_mode=off`.

This is fixed by not wrapping object files into `--start-lib/--end-lib`
for the nodeps dynamic library link action.

Fixes bazelbuild/rules_cc#230

Closes #23084.

PiperOrigin-RevId: 667945814
Change-Id: Iec148ad1b86626bb5263fecc9b98b0db89f81ad6

Commit
https://github.com/bazelbuild/bazel/commit/164ffa56dfc68151ec7f33701c304fb86a7bebda



Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>

In-memory outputs marked with the `internal-inline-outputs` execution
requirement are now added to the `inline_output_files` hint field of the
remote execution and remote cache API, potentially saving a round trip
on files that are unconditionally read by Bazel client after spawn
execution.

Support for output file inlining has been added to the remote worker
implementation.

Fixes #8421.

Closes #23428.

PiperOrigin-RevId: 673938116
Change-Id: Ieaea53502c43f9fc90b99f72dd4cdb4711239dd3 
(cherry picked from commit 56ad5d25)

Closes #23511

Users can opt into path mapping for C++ module map actions via
`--modify_execution_info=CppModuleMap=+supports-path-mapping`. This is
achieved by mapping paths in the module map files as well as converting
the sequence variable for module map paths to a new structured
`ArtifactSequenceVariable`.

Also makes it so that `ExecutionServer` gracefully handles failing
commands instead of crashing.

Closes #22957.

PiperOrigin-RevId: 675073116
Change-Id: I13835c7fb01354a89ec5fd141cf892c6b733efe4

Fixes #23178

The `runtime_solib_name` link variable had an incorrect value in the
case of a transitive dynamic library. Since non-transitive ("nodeps")
dynamic libraries are no longer used on macOS, the
`--incompatible_macos_set_install_name` flag didn't have any positive
effect.

This PR is intentionally limited to the fix so that it can be
cherry-picked into Bazel 7, where it can make the incompatible flag work
with the `apple_support` toolchain. A follow-up PR will add the feature
to the Unix toolchain and flip the incompatible flag for Bazel 8.

Work towards #12370

Closes #23089.

PiperOrigin-RevId: 668228562
Change-Id: I7524679bfe8c6b8b28c0cb04f46c0d22d0adbe99

Fixes #23185

Cherry-picks of:
* Strip unnecessary Bazel server jar contents (#22787)
* Use bootclasspath jar in proguard genrule (#23067)

Closes #23707
Closes #23708

With Xcode 16, `upb` fails its own `-Werror` check due to using Clang
extensions.

Closes #23667.

PiperOrigin-RevId: 676957876
Change-Id: I8e038b5f879077fefdb5811aff6ae5f56fb4f762

Closes #23668

PiperOrigin-RevId: 676817381
Change-Id: I66b566ed75be858b3f2d92a8ac751bc97358a578

Commit
https://github.com/bazelbuild/bazel/commit/5949aacc3a92e36e25293120c93d409f42e79b46



Co-authored-by: Googler <jhorvitz@google.com>

Also cleans up an unused field on `BazelFetchAllValue`.

Closes #23644.

PiperOrigin-RevId: 676333873
Change-Id: Ida7149a1c5642facaeed7fb68bf9f451016c359c

Commit
https://github.com/bazelbuild/bazel/commit/55346791e0a0711ef205c20bc022271c02e0c1c9



---------

Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>
Co-authored-by: Yun Peng <pcloudy@google.com>

Unquoted `query` words that start with `@@` are no longer broken on `+`,
which allows the usage of unquoted canonical labels in expressions. In
the very rare case of a shorthand form of a canonical label referring to
a WORKSPACE repo (which contains no `+`), the `+` can be separated by a
space to get the original behavior (e.g. `@@foo+bar` no longer means
`@@foo//:foo + //bar:bar`, but `@@foo +bar` does.

Closes #23600.

PiperOrigin-RevId: 673763980
Change-Id: Ie5509647b4e04e2f72b908d609781c79cf7b06d4

Fixes #23601

This regressed in
https://github.com/bazelbuild/bazel/commit/44d395338ab9a27596b0796c14076641f4cb2093.

Fixes #22980

Closes #23105.

PiperOrigin-RevId: 657276908
Change-Id: If2e88455a344082bfbec405c30c148c0d044adb6

Commit
https://github.com/bazelbuild/bazel/commit/9fbf427d3c9943f0efa659f9934e77358c64e9a7



Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>

Cherry-picks the following changes to implement output reuse:
* Deduplicate locally executed path mapped spawns (#22556)
* Fix local execution deduplication to work with optional outputs
(#23296)
* Force synchronous upload and reuse of possibly modified spawn outputs
(#23382)
* Add support for in-memory outputs to output reuse (#23422)

Fixes #23377
Fixes #23444
Fixes #23457

…a failure of the same test.

A source edit might have occurred in the interim, putting the test back
into a passing state that has been previously stored in the cache. The
current logic disables disk/remote caching when rerunning a test
following a failure, which causes spurious reexecutions (see issue
#11057).

Although this code has been around for a very long time, it seems that
it was originally added to ensure that a flaky test failure doesn't
become sticky. However, for this purpose it suffices to bust the local
cache, as a disk/remote cache should never store failures (if it did,
this would not be a complete fix, and we'd have much bigger problems
elsewhere). Note that Bazel itself never uploads local failures to a
disk/remote cache, but can't control whether a remote executor does.

The fix is to split the `TestRunnerAction.canBeCached` logic into two
versions: one that takes into account a previous test result (for the
local cache check) and another that doesn't (to determine whether we
should attempt to hit the disk/remote cache).

Fixes #11057.

PiperOrigin-RevId: 661319835
Change-Id: I9248cbfa31dd135b2bda971e48bb17d7f828889c 
(cherry picked from commit e9709b7d

)

Fixes #23259

Co-authored-by: Googler <tjgq@google.com>

Because the semantics of `--extra_toolchains` are that the last entry
has highest priority, when deduping we need to keep the **last** entry,
not the first.

Since nothing else used `dedupeOnly`, remove that entirely.

Fixes #22912.

Closes #23093.

PiperOrigin-RevId: 655696234
Change-Id: Icb5f70c47814803a1455d6b8512d977fa5f13469

Fixes #23095

Co-authored-by: John Cater <jcater@google.com>

While
https://github.com/bazelbuild/bazel/commit/cd725834eba3edb53b7a7dd9867363ef1fe00e57
made it so that a rule and an aspect can both specify validation
outputs, it accidentally didn't actually merge the `_validation` output
groups.

By adding the missing merge logic and having the `ValidateTarget` aspect
depend on all other aspects, both a rule and an aspect returning a
validation output is now fully supported with
`--experimental_use_validation_aspect`.

Closes #23589.

PiperOrigin-RevId: 675294550
Change-Id: I581417bee223f0b6aedbc0ce71cae75cd6609ef7 
(cherry picked from commit 750f0a1c)

Fixes #23664

So far only the generated header in the virtual includes is tested but
the Include path pointing to that directory isnt

This is a testcase to reproduce the suspected error reported here:
- https://github.com/bazelbuild/bazel/issues/23061

Assuming that the location of the file is asserted:
```Java
    assertThat(artifactsToStrings(ccInfo.getCcCompilationContext().getDirectPublicHdrs()))
        .contains("bin third_party/bar/_virtual_includes/starlark_lib_suffix/starlark_lib.h");
```
Adding a testcase for the include path seems reasonable.

A local run of the testcase shows the error:
```
1) testStripIncludePrefixIncludePath(com.google.devtools.build.lib.rules.cpp.StarlarkCcCommonTest)
value of: getIncludeDirs().onlyElement()
expected: bazel-out/k8-fastbuild/bin/third_party/bar/_virtual_includes/starlark_lib_suffix
but was : bazel-out/k8-fastbuild/bin/third_party/bar/_virtual_includes/starlark_lib
        at com.google.devtools.build.lib.rules.cpp.StarlarkCcCommonTest.testStripIncludePrefixIncludePath(StarlarkCcCommonTest.java:5902)

FAILURES!!!
Tests run: 73,  Failures: 1
```

The change in behaviour was introduced in
https://github.com/bazelbuild/bazel/commit/7e0df68aa4d11307dff8571ebce11d37c79c170c#diff-403c46ec3075b8e9e6d490ce955db88dae2d457b8046608884039b18b10ab6ccR774

Fixes: https://github.com/bazelbuild/bazel/issues/23061



Closes #23071.

PiperOrigin-RevId: 660681269
Change-Id: Ia3f8a8a6cf8bf0e093416a247e348e6de6719584

Closes #23221

Co-authored-by: FaBrand <21087362+FaBrand@users.noreply.github.com>

Fixes: https://github.com/bazelbuild/bazel/issues/23628

RELNOTES: Bazel fetch and vendor command now supports
--target_pattern_file for specifying target patterns.

Closes #23640.

PiperOrigin-RevId: 676063442
Change-Id: Ibbbf7879dfec4ec10093631fb002f87c9dddc8ef

Commit
https://github.com/bazelbuild/bazel/commit/3120d35eed1e9492fb848a2c4ffc7c85e7127a6a



---------

Co-authored-by: Yun Peng <pcloudy@google.com>
Co-authored-by: Xùdōng Yáng <wyverald@gmail.com>

PiperOrigin-RevId: 673946235
Change-Id: If20c63df2877aa948032d97795d88d960d9fa6bc (cherry picked from
commit ff2e950b)

…nce.

This didn't work at HEAD because the check there compared the flags in
the new configuration with the flags of the old configuration, but
overwrote the old configuration with the new one, so running a command
that discarded the analysis cache was allowed the second time.

The fix is not as trivial as moving checking the value before setting
the new configuration because creating the new configuration entails a
Skyframe evaluation, which is a waste if we end up erroring out. So the
check is now done by checking the build options of the old configuration
and the new build options, which doesn't require creating the new
configuration first.

Fixes #23491.

RELNOTES: None.
PiperOrigin-RevId: 675555324
Change-Id: Ib402b699ed9ba9f04542896a40ff1180351471db

This commit adds the `-c core.fsmonitor=false` arguments to all git
commands (excluding `git init`) in `git_worker.bzl`. The purpose of this
is to disable git from spawning a file system monitor, when many of
these are spawned on a machine that is already under load or out of file
descriptors, the git commands will hang indefinitely.

Fixes https://github.com/bazelbuild/bazel/issues/21438

Closes #23626.

PiperOrigin-RevId: 675472544
Change-Id: I79e0b6d1e406a9076cdb810747c382de52face1a

Commit
https://github.com/bazelbuild/bazel/commit/bb577b0b7c441275b83c9b8cb2ce276ec73ab2d3



Co-authored-by: Luis Padron <lpadron@squareup.com>

PiperOrigin-RevId: 670945599
Change-Id: I7aeb2c67ad3c0f17b56f4a972ec87659bd157f99

Commit
https://github.com/bazelbuild/bazel/commit/384e381fa841537987691f170ed9408dcfeed2ea



Co-authored-by: Googler <ahumesky@google.com>

PiperOrigin-RevId: 667957493
Change-Id: I013e29baa30d19ba1c5c0f1afee67b4bffd358f7

Commit
https://github.com/bazelbuild/bazel/commit/b17552bfba80e69a4d86d5566c0237e6b8438fcf



Co-authored-by: Googler <ahumesky@google.com>

This is in regards to
https://github.com/bazelbuild/bazel/commit/55535cbb9a97d6be66e41a78f66ffeee678d4fb1

[7.4.0] Ensure parent directories of action outputs are made writable when building without the bytes. (#23577)

Fixes #23462.

Closes #23555.

PiperOrigin-RevId: 672863567
Change-Id: I50af46f2ae637fb478e81ae31bd16ac4e306fe40

Commit
https://github.com/bazelbuild/bazel/commit/885a6ba3b27366a56e776f8665704eebaeb5199d



Co-authored-by: Tiago Quelhas <tjgq@google.com>

Fixes the following issues with path mapping applied to `map_each`
functions:
* Action keys could be stale if an action implementation switches from
hard-coded artifact strings to `File`s mapped at execution time. This
commit introduces a special `PathMapper` instance for fingerprinting,
see the long comment on `PathMappers.FOR_FINGERPRINTING`. It includes a
number of tests to verify that neither too much nor too little is mapped
during action key computation.
* The Starlark objects representing mapped artifact roots didn't have
the correct semantics for equality and ordered comparisons. They are now
unequal to unmapped roots.
* When not using path mapping, the `StarlarkSemantics` of a thread used
to evaluate `map_each` callbacks is not modified.

These changes require splitting `PathMappers` out of `analysis_cluster`
and also make `StrippingPathMapper` a subclass of `PathMapper` instead
of using an anonymous one.

Closes #22227.

PiperOrigin-RevId: 657449600
Change-Id: If949d23acab163cb192ea3af1baeb7a3caca7339

Closes #22227

This was missed in #19055.

Closes #23527.

PiperOrigin-RevId: 672480631
Change-Id: Ib2ef0edb6b6c90a7336085da20e2794a93763182

Commit
https://github.com/bazelbuild/bazel/commit/926b57474fbcf7ad630ebb8944b10c4b96c9f00f



---------

Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>
Co-authored-by: Xùdōng Yáng <wyverald@gmail.com>

Convert values to Starlark values when concating string_list_dict type.

This is because Dict checks that all the values put in it are Starlark
valid values. But string_list_dict value is of type List<String>, when
Java's List isn't a Starlark value.

Fixes #23065

My first PR to bazel :star_struck: Would love you to review the PR.
 Checked that the test didn't pass before the change and did after it.

Closes #23424.

PiperOrigin-RevId: 673463108
Change-Id: Ib2cdce7d94243f4e3bda23203a196fee4e766189

Commit
https://github.com/bazelbuild/bazel/commit/f1f8d5829a987854d49913058087765db68fc92c



---------

Co-authored-by: Lior Gorelik <lior10rrr@gmail.com>
Co-authored-by: Alexandre Rostovtsev <arostovtsev@google.com>

All errors used to be wrapped in an `InterruptedException` instead of
being reported with a message. Also show a warning when an incompatible
target is skipped instead of failing the build.

Work towards #21010

Closes #23210.

PiperOrigin-RevId: 660116656
Change-Id: I22651110984e471a6b31dcc59a387f3f85ad49bc


https://github.com/bazelbuild/bazel/commit/300c5867b7d2da1ba32abc20e95662096c2a7a08



Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>

Co-authored-by: Googler <ahumesky@google.com>
Co-authored-by: Googler <gnish@google.com>
Co-authored-by: Googler <cushon@google.com>
Co-authored-by: Googler <cmita@google.com>

Fixes #19055

RELNOTES: Repository rules instantiated in the same module extensions
can now refer to each other by their extension-specified names in label
attributes.

Closes #23369.

PiperOrigin-RevId: 666893202
Change-Id: Ib2eaa55fcb563adc86e16dc4a357ac808228ebda

---------

Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>

Previously, it's possible for an HTTP cache to delete CAS entries
referenced by AC without deleting the AC itself because HTTP cache
doesn't understand the relationship between AC and CAS. This could
result in permanent build errors because Bazel always trust the AC from
remote cache assuming all referenced CAS entries exist.

Now, we record the digest of lost blobs before build rewinding, so that
during the next build, we can ignore the stale AC and continue with
execution.

Fixes #18696.

RELNOTES: Added support for using a remote cache that evicts blobs and
doesn't have AC integrity check (e.g. HTTP cache).
PiperOrigin-RevId: 672536163
Change-Id: Ic1271431d28333f6d86e5963542d15a133075157

Commit
https://github.com/bazelbuild/bazel/commit/5d815790a0f5007068ed87a757e0f20aa80020ad



Co-authored-by: Googler <chiwang@google.com>

This makes it easier to realistically test how the root module would
behave as a non-root module.

RELNOTES: Overrides in the root MODULE.bazel file are now ignored with
`--ignore_dev_dependency`. (Overrides in non-root modules are already
ignored.)

Closes #23520.

PiperOrigin-RevId: 672480499
Change-Id: Ifd3a99f4ed462061101fecf918d61989c9f60401

Commit
https://github.com/bazelbuild/bazel/commit/bb7a54ceea34f901d415b096767070b0bd35a281



Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>

So that we can have proper cache eviction integration tests for HTTP
cache later.

PiperOrigin-RevId: 672501366
Change-Id: I52048a4498601cfa98cc9695d5ddddc5b93f56b9

At construction time, the action cache is not loaded so it's always
`null`. Change it to lazily get the action cache when cleaning up it.

Fixes #22220.

PiperOrigin-RevId: 672522144
Change-Id: I2de8b33ab78c04a690b17cd261d18d17f8b292ab

Commit
https://github.com/bazelbuild/bazel/commit/9187a7ec011504befba2fea98481e37558d7998c



Co-authored-by: Googler <chiwang@google.com>

The `sun.jnu.encoding` is what really determines the encoding used by
Java file system APIs (both .io and .nio) rather than `file.encoding`.
Making it more visible via an info item helps diagnose encoding bugs.

Closes #23469.

PiperOrigin-RevId: 670824910
Change-Id: Icba80b926b63090b0a5e5631e7aff673c9b71457

Commit
https://github.com/bazelbuild/bazel/commit/33ba589bfa20b850553935eca24639947787f6d5



Co-authored-by: Fabian Meumertzheim <fabian@meumertzhe.im>