| Commit message (Collapse) | Author | Age |
... | |
|
|
|
|
|
| |
ConfiguredTarget. Since ConfiguredTarget does not implement equals(), reference equality has been used until now, and this is not valid for the case where the ConfiguredTarget has been deserialized.
PiperOrigin-RevId: 199133926
|
|
|
|
|
|
| |
toolchains attribute.
PiperOrigin-RevId: 199133235
|
|
|
|
| |
PiperOrigin-RevId: 199131390
|
|
|
|
|
|
|
| |
Next step will be to skip token allocation.
RELNOTES: None.
PiperOrigin-RevId: 199121625
|
|
|
|
| |
PiperOrigin-RevId: 199118944
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 199115199
|
|
|
|
|
|
|
| |
It doesn't use CcLinkParamsStore and CppLinkAction directly anymore.
RELNOTES:none
PiperOrigin-RevId: 199107747
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This CL is a followup to c868c47. It is
intended to fix #5296. The immediate
problem in that issue is that remote caching finds a tree artifact in an input
runfiles tree without metadata in the cache. The metadata isn't there because
the runfiles artifacts are now all behind a middleman action, and
ArtifactFunction didn't propagate tree artifact values through middlemen. This
change fixes the problem by adding a tree artifact's self data value to the
values carried through to dependent actions.
Remote caching with tree artifact inputs seems to have been working mostly by
accident. (Evidence: it wasn't tested, and remote execution of actions with tree
artifact inputs results a Java traceback before and after
c868c47.) This CL is only really a bandaid to
return to the old "working" state. If remote execution wants to start supporting
tree artifacts properly in the future, we'll likely need to carry the tree
artifact child data through middleman actions, too.
Closes #5299.
Change-Id: I2e07d4fca0f6d2d34d97b7edb27f9d0063776225
PiperOrigin-RevId: 199079382
|
|
|
|
| |
PiperOrigin-RevId: 199007753
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change introduces concurrent downloads of action outputs
for remote caching/execution. So far, for an action we would
download one output after the other which isn't as bad as it
sounds as we would typically run dozens or hundreds of actions
in parallel. However, for actions with a lot of outputs or graphs
that allow limited parallelism we expect this change to positively
impact performance.
Note, that with this change the AbstractRemoteActionCache will
attempt to always download all outputs concurrently. The actual
parallelism is controlled by the underlying network transport.
The gRPC transport currently enforces no limits on the concurrent
calls, which should be fine given that all calls are multiplexed
on a single network connection. The HTTP/1.1 transport also
enforces no parallelism by default, but I have added the
--remote_max_connections=INT flag which allows to specify an upper
bound on the number of network connections to be open concurrently.
I have introduced this flag as a defensive mechanism for users
who's environment might enforce an upper bound on the number of open
connections, as with this change its possible for the number of
concurrently open connections to dramatically increase (from
NumParallelActions to NumParallelActions * SumParallelActionOutputs).
A side effect of this change is that it puts the infrastructure
for retries and circuit breaking for the HttpBlobStore in place.
RELNOTES: None
PiperOrigin-RevId: 199005510
|
|
|
|
|
|
|
|
|
|
| |
Supporting tools in inputs introduces a slow linear scan. Such tools should be moved to the 'tools' attribute. If --incompatible_no_support_tools_in_action_inputs is set the inputs are scanned, but a helpful error message is issued to the user.
Eventually we will remove the slow scanning. Errors will surface in the execution phase instead of during analysis, and the resulting error messages will be less obvious.
RELNOTES: None
RELNOTES[INC]: With --incompatible_no_support_tools_in_action_inputs enabled, Skylark action inputs are no longer scanned for tools. Move any such inputs to the newly introduced 'tools' attribute.
PiperOrigin-RevId: 198996093
|
|
|
|
|
|
|
| |
Always generating source jars can lead to action conflicts.
RELNOTES: None
PiperOrigin-RevId: 198994272
|
|
|
|
|
|
| |
Actual class to be removed in a later change.
PiperOrigin-RevId: 198937695
|
|
|
|
|
|
|
|
| |
If multiple trimming transition factories are added, they are composed in
the order they were added.
RELNOTES: None.
PiperOrigin-RevId: 198934666
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198911668
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198906931
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198906068
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198882702
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198880301
|
|
|
|
|
|
| |
nodes in ConfiguredTargetQueryEnvironemnt. This lets us support recursive target patterns with cquery.
PiperOrigin-RevId: 198879650
|
|
|
|
| |
PiperOrigin-RevId: 198877280
|
|
|
|
|
|
|
|
|
|
|
| |
java_common.compile doesn't generate the output source jar when a source jar is
the only input for the compilation. This is wrong because the source jar can
include APT generated sources. It is also inconsistent with java_library and
leads to inconsistent Skylark rules where a declared output will not always have a
generating action.
RELNOTES: None.
PiperOrigin-RevId: 198872718
|
|
|
|
|
|
|
|
| |
As the list of external repositories with their resolved information can
grow quite large, pretty print it, to make it easier for humans to inspect.
Change-Id: I34b2d6df84e52c07471a0bfaf97994cabd45feff
PiperOrigin-RevId: 198871353
|
|
|
|
|
|
| |
No longer needed since proto_library cannot be used for Java.
PiperOrigin-RevId: 198859993
|
|
|
|
|
|
| |
ActionExecutionValue.getAllFileValues.
PiperOrigin-RevId: 198807229
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
unknown commit, we are using the JavaBuilder processor exemption list instead of
disabling strict deps for the affected rules outright. This narrows the degree
to which the exemption is applied, and prevents other strict deps violations
from accruing on targets that enable databinding.
Step 2 of 2; we needed the change to roll out in JavaBuilder first to be able
to fix up issues in the depot first before removing the exemption in Blaze.
See https://github.com/bazelbuild/bazel/commit/7edc5d49c8e764ab2bdf19e3a9fb58ea30fa708f for the original attempt.
PiperOrigin-RevId: 198793976
|
|
|
|
|
|
|
|
| |
I don't know much about most of this code, but there should be no functional
changes.
RELNOTES: none
PiperOrigin-RevId: 198773657
|
|
|
|
|
|
|
| |
Setting --config in .blazerc that overrode previous options in .blazerc was triggering unnecessary warnings. Fixed by only logging warnings when a user attempts to override an option specified on the command line.
RELNOTES: None.
PiperOrigin-RevId: 198756160
|
|
|
|
|
|
|
| |
skylarkbuildapi interface is annotated.
RELNOTES: None.
PiperOrigin-RevId: 198743504
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198735582
|
|
|
|
|
|
| |
attribute.
PiperOrigin-RevId: 198735448
|
|
|
|
|
| |
RELNOTES: None.
PiperOrigin-RevId: 198728350
|
|
|
|
|
|
| |
and remove unused repositoryName field from BuildConfiguration.
PiperOrigin-RevId: 198723339
|
|
|
|
|
|
|
| |
This was a workaround created to deal with duplicated symbol errors, but there are now better ways to avoid them than using this attribute.
RELNOTES: non_propagated_deps has been removed from objc_library and apple_binary.
PiperOrigin-RevId: 198620886
|
|
|
|
|
|
| |
--experimental_one_version_enforcement_use_transitive_jars_for_binary_under_test
PiperOrigin-RevId: 198601057
|
|
|
|
|
|
|
|
| |
* is not an allowed file system character on Windows and checking if it's a
directory or file for the wilcard check would throw an InvalidPathException.
RELNOTES: None
PiperOrigin-RevId: 198595138
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
These restrictions were originally put in place to prevent things like
android_binary from being used as dependencies of android_library. However, it
doesn't actually work - the restrictions are all ORed together, and that means
rules that are of the allowed kind OR that have a JavaProvider are allowed.
Since all android_binary targets presumably have Java providers, they are
presumably all allowed.
Slightly expand these restrictions by also allowing rules that have Android
resource and asset providers. This will allow us to use Skylark rules to only
inherit assets and resources from a target, but not anything else.
I don't anticipate this expansion will have unwanted side effects, since
virtually all Android targets already have Java providers anyway.
RELNOTES: none
PiperOrigin-RevId: 198593996
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
without reading the CROSSTOOL file.
As we want to reroute selecting the toolchain from CROSSTOOL through cc_toolchain, cc_toolchain_suite should have all the necesarry information for obtaining the cc_toolchain label without accessing the CROSSTOOL file.
In order to do that, besides the existing <toolchain.targetCpu>|<toolchain.compiler>, we add <cpu> type of key in 'toolchains' attribute of cc_toolchain_suite.
Now the selection of cc_toolchain label goes as follows:
1. Return toolchains[<cpu>|<compiler> if it exists
2. Return toolchains[<cpu>] if it exists
3. Fall back to the previous state: return toolchains[<toolchain.targetCpu>|<toolchain.compiler>]
RELNOTES: None.
PiperOrigin-RevId: 198588849
|
|
|
|
| |
PiperOrigin-RevId: 198584000
|
|
|
|
|
|
|
|
|
| |
This new flag can be used to forcibly enable dSYM generation for dbg
apple builds, which is especially useful for debugging in remote
builds, which may require the use of dSYMs.
RELNOTES: Added --apple_enable_auto_dsym_dbg flag.
PiperOrigin-RevId: 198577541
|
|
|
|
|
|
|
|
| |
Fortunately or unfortunately, Skylark doesn't seem to validate against this
type, so tests don't pick this mistyping up.
RELNOTES: none
PiperOrigin-RevId: 198568282
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This used to work more smoothly. But the introduction of
ConfiguredTargetAndData adds a Precondition check:
https://source.bazel.build/bazel/+/master:src/main/java/com/google/devtools/build/lib/skyframe/ConfiguredTargetAndData.java;l=68?q=ConfiguredTargetAndData
that crashes if innerConfigurationKey is null. The only
way that can happen is when reading a bad label, which
Bazel assumes (incorrectly in this case) to be a
null-configured source file.
Good times.
PiperOrigin-RevId: 198557148
|
|
|
|
| |
PiperOrigin-RevId: 198556851
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
classes are not equal.
Also validate that there are no tree and file artifacts with the same exec path.
Fixes #4668.
Closes #5284.
Change-Id: Id97c0407a476a5bfc697b4ca7b858e3d0c0f8c75
PiperOrigin-RevId: 198540425
|
|
|
|
| |
PiperOrigin-RevId: 198535546
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
According to a post-submit regression test, this
commit increased Bazel's memory usage slightly
beyond some threshold. That event prompted me to
consider and question the necessity of this
commit.
My goal is to make Bazel work on Windows without
requiring MSYS, failing the build only if an
action is a shell action. Toolchainifying the
shell is related to, but not required by this
effort. What is required is to fail the build when
Bazel tries to create a shell action but the shell
is missing, and we have that already with
ShToolchain.getPath (which only considers the
ShellConfiguration fragment).
What's missing for my goal is to reimplement
hardcoded shell actions
(i.e. SpawnAction.builder().setShellCommand(...))
without using the shell where possible.
*** Original change description ***
shell toolchain: genrule now uses it
genrule() now uses the shell toolchain (as well as
the ShellConfiguration config fragment) to look up
the shell interpreter's path.
Also the CommandHelper no longer looks up the
shell interpreter's path itself. Instead its ctor
takes the path as an argument. This allows
supporting rules that already use the shell
toolchain and rules that still only consider the
configuration fragment.
With these changes genrule() is now able to use
the shell interpreter registered as a toolchain,
even if there's no `--shell_executable` flag
specified (or its value is empty).
Subsequent commits will migrate more and more
rules to depend on the shell toolchain.
This commit takes us closer to:
1. be able to detect the local shell interpreter's
location, especially when it's not the default
/bin/bash
2. be able to define different shell toolchains
(different interpreter paths) for remote builds
3. gracefully fail the build if the machine has no
shell installed but the action graph included a
shell action
See https://github.com/bazelbuild/bazel/issues/4319
Change-Id: I0674c7e2d5917643d87b48b19a1cb43e606ad2f7
Closes #5282.
***
RELNOTES: none
PiperOrigin-RevId: 198527351
|
|
|
|
|
| |
RELNOTES:none
PiperOrigin-RevId: 198518792
|
|
|
|
|
|
|
|
|
|
| |
I didn't catch it at the time this field was initially introduced, but R files
from Android libraries, exposed via AndroidResourcesInfo, are compiletime only.
They should not be used at runtime or inherited from dependencies. Rename the
field and add this warning to the description.
RELNOTES: none
PiperOrigin-RevId: 198448391
|
|
|
|
|
|
|
| |
targets. Along the way, update ConfiguredTargetAccessor to return the correct Target (the alias, not the actual) when dealing with AliasedConfiguredTargets.
RELNOTES: None.
PiperOrigin-RevId: 198445580
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Expose the Android data processing API to skylarkbuildapi
Rather than pass the SkylarkBuildApi interfaces (e.g., FileApi instead of
Artifact) all the way through the Android data processing code, instead just
cast them to the correct Bazel types. This is ugly, but is probably better than
having to do multiple refactorings of the Android code. (A similar refactoring,
for just a single type, took about 10 changes and 3300 lines of code.) The
Skylark Build API is only meant for things exposed to Skylark, not Bazel
internals, and, besides, when building with Bazel we should only expect inputs
from other parts of Bazel.
RELNOTES: none
PiperOrigin-RevId: 198438519
|