Previously, each thread was assigned a fixed list of work, and the main
thread would block until every thread was finished, leading to most
cores sitting idle for the last few hundred milliseconds while a few
particularly long running threads would keep working. Use a single work
queue to evenly distribute load across the threads.
Bug: http://b/32748936
Test: python run_tests.py
Change-Id: I80e231ece3b95e2721a32f658905841b89a8dc3b
Don't try to compile non-headers (vim swap files, etc.) that happen to
be in the header directory.
Bug: none
Test: manually created a file in libc/include
Change-Id: I2c1647ec7174dc617898dc2f07c9a80a08b69c72
* changes:
versioner: use a virtual filesystem for input files.
versioner: cache -cc1 flags generated by the clang Driver.
versioner: refactor to use CompilerInstance directly.
Use an InMemoryFileSystem to store and share input files across
compilations.
This improves the result of `time versioner` further, from:
versioner 109.12s user 17.43s system 2433% cpu 5.201 total
to:
versioner 112.20s user 1.38s system 2416% cpu 4.700 total
Bug: http://b/32748936
Test: python run_tests.py
Change-Id: I72d37b7c30850b8399cc40338247700fe3e7b2f9
Profiling showed that majority of time spent by versioner was being
spent in the x86 and x86_64 driver statting random files in /usr/lib,
looking for a toolchain. Hardcode a list of per-target flags which
correspond to a subset of the ones generated by clang, and use those
instead of calling out to Driver.
This changes the result of `time versioner` from:
versioner 156.57s user 1180.08s system 4109% cpu 32.525 total
to:
versioner 109.12s user 17.43s system 2433% cpu 5.201 total
Bug: http://b/32748936
Test: python run_tests.py
Change-Id: I7d254a105bf1a283cbba0546321b04e308e703d1
This will make it easier to switch over to a virtual filesystem,
which should drastically improve performance.
This also fixes an issue with warning/error reporting.
Bug: http://b/32748936
Test: python run_tests.py
Change-Id: I2e967acf75db29c95f3a03f4f94cccd22c367ad5
Clean up some output, check for errors from the versioner.
Test: tools/update_headers.sh
Bug: None
Change-Id: Ib398b534a9a8ee6dc6a2b4d90747f628aa911bab
Regenerating the NOTICE file the other day left me scratching my head at
various "how do they differ?" cases, resolved by this patch.
Test: N/A
Change-Id: I4230bfa1d7ec842a4b9fb8c73dd3b39443d74054
Generate a single NOTICE file rather than one per library. All the
headers come from libc these days anyway.
Test: tools/update_notice.sh
Bug: None
Change-Id: I127da185fdabc2815042e19aea74c507ec108f46
Also updates the license files, since apparently they're out of date.
Test: repo upload --cbr .
Bug: None
Change-Id: Ic8d855a7ee5185c4933da699292868e02ef79531
Poking around with -Wundef exposed a pretty horrifying bug, so we
should probably keep it turned on.
Bug: http://b/31496165
Change-Id: Iada279eec91ded0d59ff3841d5a3ce1a36e34c3a
Add an attribute that tells the preprocessor not to guard a
declaration, primarily for use with symbols that get reexported by
libc++ of the form `namespace std { using ::wctrans; }`.
Bug: http://b/28178111
Change-Id: I08c8751214797e37e8f26e7f7416a19e81c2bb4c
Compile headers with both -D_FILE_OFFSET_BITS=32 and -D_FILE_OFFSET_BITS=64.
Bug: http://b/30170081
Change-Id: I92651e075cc69bdc1a2581f99892c9a7fdcdb35b
Test: python run_tests.py
This was causing redundant guards to be emitted in cases where
declarations were already guarded by another macro.
Bug: http://b/30170081
Change-Id: I9080838fbea5a56225df4b26d0918657fb6c4f21
Test: python run_tests.py
Make sure everything works with relative/absolute paths, and with and
without trailing slashes.
Bug: http://b/30170081
Change-Id: Ia13a8ae2fa44590e835d0f4b006d4c31e9eaea83
Test: python run_tests.py
Move the Android.mk to src/ so that mm in that directory works.
Also, remove the dependency on FORCE_BUILD_LLVM_COMPONENTS.
Change-Id: Iad7f3fc96f5f26f9535141d752ba865b40d5ff7e
Zero initialize the contents of ArchMap, keep track of the symbol name
in Declaration, remove unnecessary copies in loops.
Change-Id: I47b07755846f252b83ffc4c89547a34b2f7ab868
Major refactor to use __attribute__((annotate)) to be able to keep
track of the semantic differences between __INTRODUCED_IN(x) and
__INTRODUCED_IN_X86(x), for use in the upcoming preprocessor.
Bug: http://b/30170081
Change-Id: I6496a8c40ba7f4553de9a2be0bbddcf37c813937
The target was moved by commit 5956b4e, but none of the headers
contained were actually necessary to compile any bionic headers.
Bug: http://b/30170081
Change-Id: I8c1a7464de0992c20af194340c0e36d131a836e7
The android-only kernel headers were moved to their own directory by
commit d6e8b8c. Add a new symlink to match.
Change-Id: I06ce7c3167fbb1e058e4a137fae9375bb5792a46
Some of the error messages emitted by versioner (the ones where it was
invoked incorrectly) go to stderr, which meant that the test runner
ignored them. Merge stdout and stderr, and switch from testing for
exact equality to endswith, because of the compilation errors test.
Change-Id: I0e2c25bcc9dea4c12ea82a6a05b29e561a61a902
Git doesn't track empty directories, so most of the tests would fail on
a fresh checkout because of dependencies/common being missing. Remove
the use of dependencies from all of the non-dependency related tests.
Change-Id: I09cc5765aae1576914c1c5d7dfa3fb666eab4a3f
