[llvm-commits] [compiler-rt] r169598 - in /compiler-rt/trunk/lib/sanitizer_common: sanitizer_common.h sanitizer_linux.cc tests/sanitizer_common_test.cc

[Dmitry Vyukov]

On Fri, Dec 7, 2012 at 10:45 PM, Kostya Serebryany <kcc at google.com> wrote:
>> > wrote:
>> >>
>> >>
>> >> On Fri, Dec 7, 2012 at 8:23 PM, Dmitry Vyukov <dvyukov at google.com>
>> >> wrote:
>> >>>
>> >>> Please check lint before committing and remove debug output from
>> >>> SanitizerCommon.SanitizerSetThreadName test.
>> >>
>> >> done, sorry. r169620.
>> >>
>> >>>
>> >>> Also, can we do something with allocator tests? Tsan tests were
>> >>> running for 1.5sec, not they run for 30sec and 29 are occupied by
>> >>> allocator tests. Are we are testing in 4 configurations.
>> >>
>> >>
>> >> I am not seeing this:
>> >> %
>> >>
>> >> ./projects/compiler-rt/lib/sanitizer_common/tests/Release/SanitizerUnitTest
>> >> [==========] 39 tests from 3 test cases ran. (4924 ms total)
>> >>
>> >> If we reduce the load in the tests we will not be catching corner cases
>> >> that
>> >> only appear under a load.
>> >>
>> >> When you run 'make -j32 check-sanitizer', gtest is called in sharded
>> >> mode,
>> >> so effectively the running time is less than 1 second.
>> >>
>> >> I guess, we are also running the allocator tests inside the tsan tests
>> >> (do
>> >> we?), which would be wrong anyway.
>> >
>> >
>> > Yes, we running all sanitizer unit tests in tsan.
>>
>> We testing both release and debug mode. I guess you run only release.
>> In tsan we test clang debug + clang release + gcc debug + gcc release.
>
>
> I see. This is the part of workflow unsupported by the cmake, so no one
> except you (and sometimes me) sees it.
> Do you really need to run the allocator (and other sanitizer_common) tests
> in those 4 modes?

Yes, I do. Debug build catches bugs for me the easy way.
Cmake build is not complete, it's only made to make minimal necessary stuff.