This project is mirrored from https://github.com/bazelbuild/bazel.git.
Pull mirroring updated .
- Aug 28, 2018
-
-
Liam Miller-Cushon authored
Fixes #5989
-
Googler authored
-
Benjamin Peterson authored
SpawnAction.Builder.setExecutable adds the executable to the inputs for you. Closes #5793. PiperOrigin-RevId: 210381105
-
jcater authored
ToolchainContextBuilder and ToolchainContext. PiperOrigin-RevId: 210377177
-
- Aug 27, 2018
-
-
Googler authored
../toolchains.html is the correct link. The other one 404s. RELNOTES: None. PiperOrigin-RevId: 210373019
-
Googler authored
RELNOTES: None PiperOrigin-RevId: 210369021
-
philwo authored
PiperOrigin-RevId: 210365824
-
Eran Shalom authored
Added test-case resolution in all log modes. https://github.com/bazelbuild/bazel/issues/1506 Closes #5429. PiperOrigin-RevId: 210362746
-
philwo authored
Because TestCase is a protobuf, the getters can never return null for String types. If the field is not set, it will return the default for the type, which is the empty string. (cf. https://developers.google.com/protocol-buffers/docs/proto#optional) The "correct" way to fix this would be to replace the null check with testCase.hasName() and testCase.hasClassName(), but by now the code hasn't worked like that for 2 years, so "fixing" it will probably do more damage by stuff invisibly relying on current behavior than simply deleting this. RELNOTES: None. PiperOrigin-RevId: 210361495
-
lpino authored
PiperOrigin-RevId: 210341075
-
Jakob Buchgraber authored
On macOS Bazel calls /usr/libexec/java_home to detect the system jdk (if any). This tool prints a warning if no system jdk is installed: > Unable to find any JVMs matching version "1.7+". > No Java runtime present, try --request to install. This patch hides this warning as Bazel ships with an embedded JDK and does not require a system JDK installed. Additionally, pump the version requirement to 1.8+ as this is the oldest JDK version that Bazel can build for. Closes #5992. PiperOrigin-RevId: 210333991
-
Googler authored
This was forgotten in https://github.com/bazelbuild/bazel/commit/52035c1c2d0ad494fb4318a6c10a6a46749ea6a2 and manifested itself in test failures (see http://b/111056578#comment5). PiperOrigin-RevId: 210318372
-
- Aug 25, 2018
-
-
Googler authored
Pass around the AndroidManifestInfo instead of the ProcessedAndroidManifest this makes the api's cleaner in Skylark. Allows us to avoid exposing non-provider objects. RELNOTES: none. PiperOrigin-RevId: 210213248
-
Googler authored
RELNOTES: none PiperOrigin-RevId: 210196034
-
Googler authored
RELNOTES: none PiperOrigin-RevId: 210193095
-
corysmith authored
RELNOTES: None PiperOrigin-RevId: 210192363
-
jingwen authored
This prevents stopping the build when invoking desugar with JDK 8. RELNOTES: None. PiperOrigin-RevId: 210172285
-
Googler authored
Possibly we could also check if the return value is empty and just set state.discoveredInputs to an empty list, but I think a skyframe restart after this point is unlikely anyway. RELNOTES: None. PiperOrigin-RevId: 210170567
-
Googler authored
RELNOTES: none. PiperOrigin-RevId: 210164466
-
Googler authored
RELNOTES: None. PiperOrigin-RevId: 210161137
-
corysmith authored
RELNOTES: None PiperOrigin-RevId: 210160726
-
tomlu authored
This is controlled by a flag "expand_directories", whose default value is controlled by --incompatible_expand_directories. Example: directory = ctx.actions.declare_directory('dir') ... args = ctx.actions.args() args.add_all([directory]) The resulting command line will now contain the directory contents instead of the path to the directory. RELNOTES[INC]: Adds --incompatible_expand_directories to automatically expand directories in skylark command lines. Design doc: https://docs.google.com/document/d/11agWFiOUiz2htBLj6swPTob5z78TrCxm8DQE4uJLOwM PiperOrigin-RevId: 210155043
-
ccalvarin authored
Rollforward of https://github.com/bazelbuild/bazel/commit/2d3adfb93c023f9752f00d4bf3440ee9869f64f2: Add a warning if the same rc file is read multiple times. Prerequisite for fixing #5765, to minimize the risk that multiple rc files import the same file, which is slightly more likely with try-import. Added extra test for double-load rcs imported using different syntax, fixing an issue where paths were not canonicalized for reliable comparison. RELNOTES: None. PiperOrigin-RevId: 210140725
-
Googler authored
RELNOTES: None PiperOrigin-RevId: 210121612
-
Googler authored
RELNOTES: None. PiperOrigin-RevId: 210117999
-
Googler authored
java.util.HashMap.get() has an optimization where it compares the hashcode to a cached value before calling equals() on the candidate key. CompactHashSet/Map keep a table of cached hashcodes. But RegularImmutableMap only has a comment saying it assumes the objects in the collection will optimize equals() with == and hashcode as appropriate! Simplify use of Objects.equal(); none of the fields are nullable. Take getClass() out of hashcode computation. I think two different Artifact subclasses for the same path is unusual in normal operation, and I may try to exploit the pass-through hashcode in ActionInputMap. RELNOTES: None. PiperOrigin-RevId: 210117899
-
laurentlb authored
RELNOTES: None. PiperOrigin-RevId: 210117808
-
lberki authored
Also add a few stern warnings about the extreme badness of the idea of handling a Path object in the analysis phase. RELNOTES: None. PiperOrigin-RevId: 210115418
-
jsharpe authored
When building remotely on Windows, use the Java version of singlejar, not the native one, since the native one doesn't work yet. Work to make the native version work tracked at https://github.com/bazelbuild/bazel/issues/2241 This is needed to run singlejar for remote builds on Windows successfully. RELNOTES: N/A PiperOrigin-RevId: 210105718
-
- Aug 24, 2018
-
-
lberki authored
And a little bit of associated refactoring. RELNOTES: None. PiperOrigin-RevId: 210097491
-
Klaus Aehlig authored
...into a computation of the Skylark value contained in a resolved file, and the computation of the expected hashes from this value. In this way, we can reuse the code for computing the resolved value for the (experimental) feature of using a resolved file instead of a workspace file. PiperOrigin-RevId: 209150452 Change-Id: Ib1a77128100d821c4cc3e0a0ba57f65d44e07e97 PiperOrigin-RevId: 210085624
-
corysmith authored
Switch all dummy manifest writing to the AndroidManifest class. This ensures that the minSdk will be propagated to the dummy manifests. RELNOTES: None PiperOrigin-RevId: 210085414
-
lberki authored
Apparently, this is https://bugs.openjdk.java.net/browse/JDK-8056066 triggered by https://github.com/bazelbuild/bazel/commit/deccc485603c004daad959fd747f1c0c9efc4f00 (fixed in http://openjdk.java.net/jeps/216 but that doesn't help JDK 8) RELNOTES: None. PiperOrigin-RevId: 210078552
-
lberki authored
Also remove ProfileArtifacts because it was a useless abstraction. RELNOTES: None. PiperOrigin-RevId: 210078489
-
Ola Rozenfeld authored
Change-Id: I245c8524adedfb343e3ffd41a54714f40fc193c4
-
juliexxia authored
This allows for properly resolved selects for the proto output format of cquery (b/112853400). AliasConfiguredTargets don't extend RuleConfiguredTargets, just stores an 'actual' configured target value. Since they don't extend RuleConfiguredTarget, they don't have access to the config conditions for the alias target. They *can* currently access the config condition for the 'actual' target, but those aren't the correct conditions for the alias target. ex of pattern that this CL addresses: //test:BUILD alias( name = 'fake_target', actual = select({ ':config1': ':target1', ':config2': ':target2', }) ) Currently, running "blaze cquery 'deps(//test:fake-target)' --output=proto --correct_flag_to_trigger_config1" returns an error since the output formatter attempts to use the config conditions of target1 (the resolved value of the 'actual' attribute) to evaluate the select in the 'actual' attribute of the alias target. PiperOrigin-RevId: 210002631
-
corysmith authored
RELNOTES: None PiperOrigin-RevId: 210002035
-
ccalvarin authored
RELNOTES: None. PiperOrigin-RevId: 209999857
-
Googler authored
RELNOTES: None. PiperOrigin-RevId: 209995144
-
corysmith authored
Stop passing the merged xml resources to the resource shrinker. RELNOTES: None PiperOrigin-RevId: 209989758
-