| Commit message (Collapse) | Author | Age |
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143972504
MOS_MIGRATED_REVID=143972504
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
== Before this change ==
Previously, if a and b are sets, then a + b created a new set c whose direct and transitive elements were all those of a, and with b appended as an additional transitive element. If on the other hand b is a list instead of a set, then its contents were appended as additional direct elements of c. In both cases, you can think of c as a copy of a that also knows about b.
This copying of a's elements into c can lead to accumulation when you do it repeatedly, e.g. x += y in a loop. Each union can take O(n) time so you get O(n^2) time overall. Nested set union is supposed to be O(1) time per operation and O(n) time overall.
It also leads to surprising iteration orders. If you do a + b + c + d (left-associative), where each one is a set, then when you do a post-order traversal you get the elements of b, c, d, and a, in that order. This is because b, c, and d each get appended as transitive children of the copies of a.
== After this change ==
If a and b are sets, then a + b returns a new set c with a and b as its two transitive children. If b is a list, then c has a as its only transitive child and b's elements as its only direct elements. This is straightforward, O(1), and avoids the problem with the confusing order.
It is implemented by removing the items/transitiveItems fields and just relying on NestedSetBuilder.
RELNOTES[INC]: (Skylark) Set union is now O(1). As a side-effect, the iteration order of sets produced by union has changed. "print(set([1]) + set([2]) + set([3]))" will now give back the order 1, 2, 3 instead of 2, 3, 1.
--
PiperOrigin-RevId: 143972165
MOS_MIGRATED_REVID=143972165
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes #2311
Closes #2312.
--
Reviewed-on: https://github.com/bazelbuild/bazel/pull/2312
PiperOrigin-RevId: 143967880
MOS_MIGRATED_REVID=143967880
|
|
|
|
|
|
|
|
| |
The skylark provider is bound as "proto_java" to avoid collisions with the base, which is called "java".
--
PiperOrigin-RevId: 143960605
MOS_MIGRATED_REVID=143960605
|
|
|
|
|
|
|
|
|
|
| |
This feature allows us to expand a flag_group when a build variable is not
available. This is helpful when migrating crosstools in a backward compatible
way (that works with released bazel as well as with bazel at HEAD).
--
PiperOrigin-RevId: 143955333
MOS_MIGRATED_REVID=143955333
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
*** Reason for rollback ***
Breaks //src/test/shell/bazel:bazel_sandboxing_test:
http://ci.bazel.io/job/bazel-tests/lastCompletedBuild/BAZEL_VERSION=latest-jdk7,PLATFORM_NAME=linux-x86_64/testReport/(root)/(empty)/test_failing_action_with_ioexception_while_copying_outputs_throws_correct_exception/
*** Original change description ***
Gives 3 levels of sandbox error message under different flags.
1. no flag: only the direct reason is given (command execution termination status), and also the instruction to use "--verbose_failures"
2. flag "--verbose_failures": gives failed execution command and the instruction to use "--sandbox_debug --strategy"
3. flag "--verbose_failures --sandbox_debug": gives failed execution command, debugging message from sandboxing, and the instruction to use "--strategy"
Also removes "cd <sandbox_...
***
ROLLBACK_OF=143937589
--
PiperOrigin-RevId: 143951901
MOS_MIGRATED_REVID=143951901
|
|
|
|
|
|
|
|
|
| |
We were previously jumping through hoops to avoid cyclic dependencies. It's
unclear why the class was here in the first place rather than in lib.actions.
--
PiperOrigin-RevId: 143948959
MOS_MIGRATED_REVID=143948959
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143939410
MOS_MIGRATED_REVID=143939410
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
1. no flag: only the direct reason is given (command execution termination status), and also the instruction to use "--verbose_failures"
2. flag "--verbose_failures": gives failed execution command and the instruction to use "--sandbox_debug --strategy"
3. flag "--verbose_failures --sandbox_debug": gives failed execution command, debugging message from sandboxing, and the instruction to use "--strategy"
Also removes "cd <sandbox_path>" in given failed command, since debugging is only necessary with flag "--verbose_failures --sandbox_debug" and the path is already given in sandboxing debugging message.
Addresses #2174.
--
PiperOrigin-RevId: 143937589
MOS_MIGRATED_REVID=143937589
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143936734
MOS_MIGRATED_REVID=143936734
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143933349
MOS_MIGRATED_REVID=143933349
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143823562
MOS_MIGRATED_REVID=143823562
|
|
|
|
|
|
|
|
| |
extension registry for the enclosing binary.
--
PiperOrigin-RevId: 143807030
MOS_MIGRATED_REVID=143807030
|
|
|
|
|
|
|
|
| |
The main effect is to cause cc_test rules to build into a self-contained executable rather than relying on external shared libraries. (cc_binary rules do this by default already.)
--
PiperOrigin-RevId: 143806513
MOS_MIGRATED_REVID=143806513
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143792017
MOS_MIGRATED_REVID=143792017
|
|
|
|
|
|
|
|
| |
changes are visibile to all other threads by volatility and we don't read the variable at any intermediate or inconsistent state (simply check against null or a constant). reverseDepsToSignal is also made volatile for subclasses that need volatile reads to it.
--
PiperOrigin-RevId: 143787032
MOS_MIGRATED_REVID=143787032
|
|
|
|
|
|
|
|
| |
outside of rule implementation.
--
PiperOrigin-RevId: 143720112
MOS_MIGRATED_REVID=143720112
|
|
|
|
|
|
|
|
| |
and 3 different binary types.
--
PiperOrigin-RevId: 143718198
MOS_MIGRATED_REVID=143718198
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
TESTED=
$ ls android-ndk-r12b/platforms
android-12 android-15 android-18 android-22 android-9
android-13 android-16 android-19 android-23
android-14 android-17 android-21 android-24
RELNOTES: Removed broken api levels 3, 4 and 5 from Android NDK 12.
--
PiperOrigin-RevId: 143710544
MOS_MIGRATED_REVID=143710544
|
|
|
|
|
|
|
|
|
|
|
| |
Since we evaluate the BUILD file despite errors (e.g. parse errors), we
should provide the .bzl dependencies to the BUILD file.
This change removes some confusing/irrelevant error messages.
--
PiperOrigin-RevId: 143696291
MOS_MIGRATED_REVID=143696291
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(Tests kept.)
*** Reason for rollback ***
[]
*** Original change description ***
When --experimental_oom_more_eagerly is enabled, tell Bazel to exit with an OutOfMemoryError and have the JVM send Bazel a SIGUSR2 when it detects an OOM. This should help in certain pathological cases when Bazel GC thrashes for some time after an OOM has been detected.
--
PiperOrigin-RevId: 143694970
MOS_MIGRATED_REVID=143694970
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143694257
MOS_MIGRATED_REVID=143694257
|
|
|
|
|
|
|
|
|
|
| |
This are random little nits that aren't dependent on changing the
directory structure, so it makes the (rather large) CL that is
coming cleaner.
--
PiperOrigin-RevId: 143690681
MOS_MIGRATED_REVID=143690681
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
android_ndk_repository.
Read from ANDROID_{NDK_}HOME if not set. Note that the repository is NOT invalidated if ANDROID_{NDK_}HOME is changed. Once https://bazel.build/designs/2016/10/18/repository-invalidation.htm is implemented, that will no longer be the case.
This is one piece of https://github.com/bazelbuild/bazel/issues/2284.
RELNOTES: android_sdk_repository and android_ndk_repository now read $ANDROID_HOME and $ANDROID_NDK_HOME if the path attribute is not set.
--
PiperOrigin-RevId: 143686964
MOS_MIGRATED_REVID=143686964
|
|
|
|
|
|
|
|
|
| |
The code assumed that the last element before a cycle was a BUILD file. It can
also be a .bzl file.
--
PiperOrigin-RevId: 143673940
MOS_MIGRATED_REVID=143673940
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143666208
MOS_MIGRATED_REVID=143666208
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143664295
MOS_MIGRATED_REVID=143664295
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143601292
MOS_MIGRATED_REVID=143601292
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This merges the AndroidResourceContainerBuilder (which it's not even clear is related to
the nested ResourceContainer!) into the newly generated ResourceContainer.Builder.
It also seemed ridiculous for ResourceContainer to get so large and still be subordinate
to AndroidResourcesProvider, especially when it's getting passed around in a lot of other
places (look how many imports needed fixing!). This CL makes it its own top level class.
This allows for easy modification of an existing instance: call toBuilder on it, set
the properties you want set, and then call build.
--
PiperOrigin-RevId: 143574468
MOS_MIGRATED_REVID=143574468
|
|
|
|
|
|
|
|
|
|
| |
See OutputFileConfiguredTargetTest for deep details.
Also more strongly enforce the expectation that all output files have generating rules.
--
PiperOrigin-RevId: 143570028
MOS_MIGRATED_REVID=143570028
|
|
|
|
|
|
|
|
| |
to only contain direct dependencies.
--
PiperOrigin-RevId: 143557048
MOS_MIGRATED_REVID=143557048
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143555554
MOS_MIGRATED_REVID=143555554
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143547986
MOS_MIGRATED_REVID=143547986
|
|
|
|
|
|
|
|
| |
In case of conflict, show the location of the original rule.
--
PiperOrigin-RevId: 143541281
MOS_MIGRATED_REVID=143541281
|
|
|
|
|
|
|
|
|
|
| |
solved.
--
Change-Id: Ie9314cabda8b5542979f58cdb2ab7a41e6cd26ad
Reviewed-on: https://cr.bazel.build/8091
PiperOrigin-RevId: 143470985
MOS_MIGRATED_REVID=143470985
|
|
|
|
|
|
|
|
| |
--
Change-Id: I9ec3209a69ba5a51943b334f278ba93d67d4f9f4
Reviewed-on: https://cr.bazel.build/8090
PiperOrigin-RevId: 143470915
MOS_MIGRATED_REVID=143470915
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143454933
MOS_MIGRATED_REVID=143454933
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
instead of getting them from the base.
This is a preparatory change to allow aspects to bind JavaSkylarkApiProvider with a different name.
java_proto_library needs to be able to advertise its jars from each aspect. The elegant way to do this would be to bind the "java" skylark provider. Unfortunately, proto_library also advertises this provider, even though it is incorrect to do so in the context of java_proto_library, because the jars advertised by the base configured target are certainly not going to be produced.
Unless we can override providers when the merged configured target is produced, we then need to be able to bind a java provider with a different name. Unfortunately the implementation of this accesses providers from the configured target, which again belong to the proto_library, not the aspect. This CL allows the java skylark api provider to serve data defined at construction time rather than from hard-coded providers from the configured target.
--
PiperOrigin-RevId: 143445977
MOS_MIGRATED_REVID=143445977
|
|
|
|
|
|
|
|
|
|
|
| |
compiling/linking instead of --watchos_sdk_version.
RELNOTES: Set clang's -mwatchos-version-min correctly using the value of
--watchos_minimum_os, not --watchos_sdk_version.
--
PiperOrigin-RevId: 143444638
MOS_MIGRATED_REVID=143444638
|
|
|
|
|
|
|
|
| |
Remove build flag --experimental_j2objc_annotation_processing. It is on by default now.
--
PiperOrigin-RevId: 143444299
MOS_MIGRATED_REVID=143444299
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143443361
MOS_MIGRATED_REVID=143443361
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
TL;DR: resolve "C:/progra~1" style paths, don't
cache failed resolutions in the parent's
Path.children
When creating WindowsPath objects, resolve the
"C:/progra~1" style paths to "C:/program files".
This enables us to correctly compare paths on
Windows. Without this canonicalization we
incorrectly determine "C:/progra~1" and
"C:/program files" to be different when they are
actually the same.
We only attempt to resolve such paths if the name
looks like it's an abbreviated path.
If resolution fails, probably due to the path not
existing, then we don't cache this Path object in
the parent's `children` list. This avoids stale
cache entries in case the path springs into
existence later in the server's lifetime.
Fixes https://github.com/bazelbuild/bazel/issues/2145
We also need to rectify https://github.com/bazelbuild/bazel/issues/2173
--
PiperOrigin-RevId: 143442134
MOS_MIGRATED_REVID=143442134
|
|
|
|
|
|
|
|
| |
platform-default and latin-1
--
PiperOrigin-RevId: 143439199
MOS_MIGRATED_REVID=143439199
|
|
|
|
|
|
|
|
|
| |
This will be used by LibrariesToLinkValue to switch on many different types of
libraries.
--
PiperOrigin-RevId: 143438434
MOS_MIGRATED_REVID=143438434
|
|
|
|
|
|
|
|
| |
The start time of critical path component wasn't recorded correctly since #getStartWallTimeMillis() didn't work right. However, since profiler tasks store nanotime anyway, it's pointless to try to convert back and forth.
--
PiperOrigin-RevId: 143438289
MOS_MIGRATED_REVID=143438289
|
|
|
|
|
|
|
|
| |
Fixes #2325.
--
PiperOrigin-RevId: 143434786
MOS_MIGRATED_REVID=143434786
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143434756
MOS_MIGRATED_REVID=143434756
|
|
|
|
|
|
|
|
| |
CcToolchainFeatures.FlagGroup constructor
--
PiperOrigin-RevId: 143429631
MOS_MIGRATED_REVID=143429631
|
|
|
|
|
|
| |
--
PiperOrigin-RevId: 143390220
MOS_MIGRATED_REVID=143390220
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
strip(), lstrip(), and rstrip() now accept a None argument as a synonym for
the no-arg version. The characters that are considered as whitespace (to be
removed by default in the no-arg form) are now the same as in Python 3.6.
RELNOTES[INC]: The string methods strip(), lstrip(), and rstrip() now
by default remove the same whitespace characters as Python 3 does, and accept
None as an argument.
--
PiperOrigin-RevId: 143388250
MOS_MIGRATED_REVID=143388250
|