- 06 Apr, 2022 2 commits
- 13 Jan, 2022 4 commits
-
- 10 Jan, 2022 2 commits
-
-
Branislav Rankov authored
The write_source_info got broken in the case when there is no PATCHES.json file. Skipped listing patches if this file is missing. Change-Id: If531120c9d8d2dbd4504a2ad0e08e3d1c1c39b3e
-
Branislav Rankov authored
This merges the revision used in aosp toolchain r433403 into the morelo/master branch. Conflicts: android_version.py base_builders.py do_build.py paths.py source_manager.py trim_patch_data.py Change-Id: If49ef4425d027ebc7565d94f02760a9c71dc07e5
-
- 19 Oct, 2021 1 commit
-
-
Chih-Hung Hsieh authored
* fix/suppress some but not all existing pylint warnings * remove some unused imports * reorder some imports * add a missing timer import, which is needed by --run-tests-stage1 * add "bugs_tests" to PATCHES.json to keep any number of comment strings for a patch. * updated/renewed patches: * Revert-Sema-Introduce-BuiltinAttr-per-declaration-bu.v4.patch (with 14 clang tests marked as XFAIL) * pick upstream bug fixes: * [OpenMP][libomptarget] Change device vector elements to unique_ptr type * Revert [Coroutines] [Clang] Look up coroutine component in std namespace first", * Add missing `REQUIRES: asserts` to combine-icmp-to-lhs-known-bits.mir * [MergeICmps] Don't reorder unmerged comparisons * builders.py config changes: * libclang_rt.builtins-*-android.a moved from 'android' to 'linux'. * defines['LIBUNWIND_TARGET_TRIPLE'] * defines['LIBCXXABI_TARGET_TRIPLE'] * defines['LLDB_ENABLE_LUA'] = 'OFF' * copy omp.h and omp-tools.h to the clang include directory * configs.py changes: * def llvm_triple * do_build.py config changes: * getVersions(libname) * soname_version Bug: 201432564 Test: build.py --no-build=windows --run-tests-stage1 Change-Id: I44411cead4c6237ab4dc8fa8f62d54e4e48c0767
-
- 18 Oct, 2021 2 commits
-
-
Pirama Arumuga Nainar authored
We built it for the host to get openmp headers. Instead copy headers from one of the device builds. Test: build.py and verify absence of omp host libs and presence of omp headers. Change-Id: I7f3c3e2f089e3573e55e6526edd856d1e75f5c67
-
Pirama Arumuga Nainar authored
Bug: http://b/199315634 69708477be2 [compiler-rt/profile] Hide __llvm_profile_raw_version Test: N/A Change-Id: Iffd90f180ff15ce26e80a12072171fca7378670a
-
- 14 Oct, 2021 1 commit
-
-
Stephen Hines authored
Bug: http://b/203003319 Test: ./build.py Change-Id: I3f467ea4908f6e9e84a86efef160002639eaa082
-
- 13 Oct, 2021 3 commits
-
-
Stephen Hines authored
Bug: http://b/201691289 Test: N/A Change-Id: Ia300c1dd71a5eeca492ff475211349d475945dfe
-
Chih-hung Hsieh authored
-
Chih-Hung Hsieh authored
* --skip-tests is passed to build.py * multiple --module options can be used to replace the default 'dist' Test: toolchain/llvm_android/test_compiler.py --build-only --skip-tests \ --module tidy-external-tinyxml2 --module tidy-system-extras-simpleperf \ --with-tidy --target aosp_blueline-eng . Change-Id: Icdec10e0bbc077e8f8dc58dce3f7fc9df3e3b198
-
- 12 Oct, 2021 1 commit
-
-
Nick Desaulniers authored
Bug: http://b/201029275 85f612efeb3 DebugInfo: Use sugared function type when emitting function declarations for call sites Test: N/A Change-Id: I46654c80ef6d0c25d78ff1066776d268ee6e873a
-
- 09 Oct, 2021 1 commit
-
-
Yi Kong authored
This is unnecessary and make PGO collection pipeline slower. Besides, this could taint the 3-stage CSIR PGO profile (once it lands). Test: test_compiler.py --build-only --target aosp_flame-eng --generate-clang-profile ./ Change-Id: I6fab211e0a3eb0cfc53e9cb94ee75d9b6ba089da
-
- 07 Oct, 2021 1 commit
-
-
Treehugger Robot authored
-
- 06 Oct, 2021 1 commit
-
-
Pirama Arumuga Nainar authored
Bug: http://b/202303953 Test: N/A Change-Id: I5c334137caf08acd32c53ed0070cd63594158972
-
- 05 Oct, 2021 2 commits
-
-
Stephen Hines authored
Bug: http://b/202099787 Test: ./build.py Change-Id: I9fc04eca9f8ea9f97f7b06f7cb5fd5cc3807f3ed
-
Stephen Hines authored
Bug: http://b/202099787 31e75512174 [DebugInfo] Correctly update debug users of SSA values in tail duplication Test: N/A Change-Id: I5beaf5e1430028c502e94e0c2b3af92cf1a33cbc
-
- 30 Sep, 2021 2 commits
-
-
Treehugger Robot authored
-
Pirama Arumuga Nainar authored
This reverts commit c03774b4. Reason for revert: publish artifacts can now support <SHA.kzip> Original commit message: The SHA will be used by the kythe backend to match crossreferences to the source. Currently, the build will fail if the SHA is not available on the build server. We can add a fallback or alternative way of getting the SHA if the build failures are too noisy. Bug: http://b162557156 Change-Id: Ic0a3ddd627f7f4692b2991e743153e932d2326ff Test: toolchain/llvm_android/kythe_xref.py
-
- 22 Sep, 2021 2 commits
-
-
Chih-Hung Hsieh authored
The test should use the default checks specified in build/soong/cc/config/tidy.go. Test: test_compiler.py --build-only --target aosp_blueline-eng ./ Bug: 200851002 Change-Id: Ie467dbb79d3aaec9ee3e0103da793b8b0c9f71e1
-
Yi Kong authored
Bug: http://b/199471467 d0746f2e9bb Don't fold (select C, (gep Ptr, Idx), Ptr) if C is vector but Idx is scalar Test: N/A Change-Id: I2337f109326b2641a47c89afe181f5d32f494f24
-
- 21 Sep, 2021 1 commit
-
-
Pirama Arumuga Nainar authored
This covers config + build steps as well as the test step. Test: build.py and inspect $DIST_DIR/build_times.txt Change-Id: I4276208b9c21ea5596701b81b01f1bb4afd921d1
-
- 17 Sep, 2021 2 commits
-
-
Treehugger Robot authored
-
Pirama Arumuga Nainar authored
This policy is on by default after CMake 3.2. OpenMP's minimum version is now higher than 3.2 so explicit opt in is no longer necessary. Test: build.py Change-Id: If4f034b733ba59264e28ba3936a07492906946bd
-
- 16 Sep, 2021 2 commits
-
-
Pirama Arumuga Nainar authored
Bug: http://b/162557156 Using the vname mapping removes the toolchain/llvm-project prefix from filenames. Test: kythe_xref.py and inspect a file under 'root/units/' in the output kzip. Change-Id: I4db90cfb206eb76c3b9c7876face4281b7c3139c
-
Kevin Brodsky authored
Since c3112d89 ("Use git_llvm_rev.py to compute SVN revision"), _get_llvm_remote_and_branch_info() inspects all remote branches to find the "best match" upstream branch. This operation can take quite a bit of time, and we can avoid it. Indeed, if we are on a branch that has an upstream branch, then we can simply use that branch. Otherwise, it is reasonable to fall back to the remote branch configured in the repo manifest (accessible via the "fake" `m` remote). Another modification this change introduces is that we no longer require the upstream branch to be behind (or equal to) HEAD. This is not necessarily the case after a `git fetch` (for instance). Instead, we simply count the number of commits on top of the upstream branch, disregarding the number of commits we are behind. Change-Id: I910ed71ae84cdd4f96819f41b5d01fa4d1a13769
-
- 14 Sep, 2021 1 commit
-
-
Chih-Hung Hsieh authored
Test: build.py --no-build windows --run-tests-stage1 Change-Id: I16cfcc524df5c057933be01a36a8f3f188a2f4b4
-
- 13 Sep, 2021 1 commit
-
-
Yi Kong authored
Test: presubmit Change-Id: I7b6ee1957833bb204cf33a4e03a58d469b298875
-
- 09 Sep, 2021 3 commits
-
-
Dan Albert authored
Bug: http://b/175635923 d280a76908e [Driver][Linux] Fix regression when -DLIBCXX_LIBDIR_SUFFIX=64 Test: N/A Change-Id: If3d4a93ab9a92b864e9b1d866b37a9ca0d958f4d
-
Treehugger Robot authored
-
Dan Albert authored
No failures :) I haven't enabled the libc++abi tests because they don't have the equivalent cxxflags/ldflags options that libc++ does so I can't tell the test runner how to find libgcc and lld. That should be easy to solve, but we're also on the legacy (default) test runner right now and I think the new one simplifies all that anyway. Test: treehugger Bug: http://b/175635923 Change-Id: I57906f3e6606665f8f853b6035ee84db91475814
-
- 08 Sep, 2021 1 commit
-
-
Pirama Arumuga Nainar authored
The script previously assumed that the HEAD of toolchain/llvm-project points to the last merge commit but that's incorrect. We could have local changes (e.g. to OWNERS files). Use `git log --merges -n 1` instead. Test: Inspect out/clang_source_info.md Change-Id: I6e09e17992726836099163325641379c85b1c57c
-
- 05 Sep, 2021 1 commit
-
-
Pirama Arumuga Nainar authored
Bug: http://b/162557156 Bug: http://b/198867707 Test: N/A Change-Id: I9dc193ea1019c8d82b60d63db51df133f5191b9e
-
- 03 Sep, 2021 1 commit
-
-
Pirama Arumuga Nainar authored
This reverts commit 46e4ab67. Reason for revert: publish_artifacts cannot upload <SHA>.kzip yet. Change-Id: Ic76246b80eb92a7ef5e5c9aac24831200ec4afdb
-
- 30 Aug, 2021 2 commits
-
-
Pirama Arumuga Nainar authored
The SHA will be used by the kythe backend to match crossreferences to the source. Currently, the build will fail if the SHA is not available on the build server. We can add a fallback or alternative way of getting the SHA if the build failures are too noisy. Bug: http://b162557156 Test: toolchain/llvm_android/kythe_xref.py Change-Id: Ic349c1918922897b0f7891e26df8637906cde4eb
-
Treehugger Robot authored
-