[llvm] r292862 - [sanitizer-coverage] emit __sanitizer_cov_trace_pc_guard w/o a preceding 'if' by default. Update the docs, also add deprecation notes around other parts of sanitizer coverage

Tue Jan 24 17:27:28 PST 2017

Fixed and re-enabled the test in r292998.
Will watch the bots.

Sorry for the breakage.
The bug was pretty subtle and there was no way for me to debug the PPC
failure.
Luckily the gcc build also broke, so I had a repro.

--kcc

On Tue, Jan 24, 2017 at 4:10 PM, Chandler Carruth <chandlerc at google.com>
wrote:

> Thanks!
>
> On Tue, Jan 24, 2017 at 4:09 PM Kostya Serebryany <kcc at google.com> wrote:
>
>> r292994 disables the failing parts of the test, still investigating.
>>
>> On Tue, Jan 24, 2017 at 3:56 PM, Kostya Serebryany <kcc at google.com>
>> wrote:
>>
>> yes, give me 30 minutes.
>>
>> On Tue, Jan 24, 2017 at 3:50 PM, Chandler Carruth <chandlerc at google.com>
>> wrote:
>>
>> So, we've lost over a day of build bot coverage on PPC.
>>
>> Kostya, before detailed investigation, can you at least let the bots
>> return to green? I don't really care between revert and disable on PPC. I
>> just want these bots to go green:
>> http://lab.llvm.org:8011/builders/clang-ppc64be-linux-lnt
>> http://lab.llvm.org:8011/builders/clang-ppc64le-linux-lnt
>> http://lab.llvm.org:8011/builders/clang-ppc64be-linux-multistage
>> http://lab.llvm.org:8011/builders/clang-ppc64le-linux-multistage
>> http://lab.llvm.org:8011/builders/clang-ppc64be-linux
>> http://lab.llvm.org:8011/builders/clang-ppc64le-linux
>>
>>
>> On Tue, Jan 24, 2017 at 3:48 PM Kostya Serebryany via llvm-commits <
>> llvm-commits at lists.llvm.org> wrote:
>>
>> turns out one of the x86 bots is failing too (gcc build fails, clang
>> build passes).
>> http://lab.llvm.org:8011/builders/sanitizer-x86_64-
>> linux/builds/721/steps/check-asan%20in%20gcc%20build/logs/stdio
>> This gives me a way to reproduce, looking.
>>
>>
>> On Tue, Jan 24, 2017 at 1:43 PM, Eric Christopher <echristo at gmail.com>
>> wrote:
>>
>>
>>
>> On Tue, Jan 24, 2017 at 1:37 PM Kostya Serebryany <kcc at google.com> wrote:
>>
>> On Tue, Jan 24, 2017 at 1:34 PM, Eric Christopher <echristo at gmail.com>
>> wrote:
>>
>>
>>
>> On Tue, Jan 24, 2017 at 1:23 PM Kostya Serebryany <kcc at google.com> wrote:
>>
>> On Tue, Jan 24, 2017 at 1:17 PM, Eric Christopher <echristo at gmail.com>
>> wrote:
>>
>>
>>
>> On Tue, Jan 24, 2017 at 11:49 AM Kostya Serebryany <kcc at google.com>
>> wrote:
>>
>> On Tue, Jan 24, 2017 at 11:45 AM, Eric Christopher <echristo at gmail.com>
>> wrote:
>>
>> This seems problematic in general - we have a bot that is failing due to
>> a particular test. Is this patch necessary for correctness somewhere else?
>>
>>
>> Correctness -- no. Performance -- yes, and in some cases very significant.
>>
>>
>> OK. Thanks.
>>
>>
>>
>>
>> Can it be reverted until we can figure out what's going on?
>>
>>
>> Sure we can revert it, but who is going to analyze the problem and how
>> soon?
>>
>>
>> I'm hoping someone at IBM can take a look at it. I've got limited ability
>> to debug it myself at the moment.
>>
>> Kit: Would you mind terribly?
>> Kostya: It might be nice to revert in the mean time if you don't mind?
>>
>>
>> I do mind, sorry.
>> Again, this functionality is not used by anyone on PPC (afaict) but is
>> important for libFuzzer users on x86 side.
>>
>>
>> I believe this could be causing problems not just in libFuzzer and so
>> affect anyone trying to use the sanitizer on ppc at all?
>>
>>
>> I doubt so. This change affects sanitizer coverage (
>> https://clang.llvm.org/docs/SanitizerCoverage.html),
>> which is usually used together with *san and libFuzzer.
>>
>>
>> OK. Thanks for the clarification.
>>
>> In general, we have a revert first policy and so this should probably
>> have fallen under there without letting people know what the plan was for
>> making sure the bots come back - it's important that we continue to get bot
>> testing for asan in general.
>>
>> Thanks!
>>
>> -eric
>>
>>
>>
>>
>>
>> If not, can you explain the scope of this a bit better - especially as
>> it's now a flaky failure?
>>
>> -eric
>>
>>
>>
>>
>>
>> Thanks!
>>
>> -eric
>>
>>
>> Note that the functionality tested in this test is not the core asan, but
>> the coverage instrumentation used by libFuzzer,
>> and afaict there are no users of libFuzzer on PPC, so we can just disable
>> the test on PPC and no users will be harmed.
>>
>>
>>
>> -eric
>>
>> On Tue, Jan 24, 2017 at 11:29 AM Kostya Serebryany via llvm-commits <
>> llvm-commits at lists.llvm.org> wrote:
>>
>> Hi Bill,
>>
>> thanks for the report.
>> Unfortunately, I don't have anyone working on the PPC side at the moment
>> and the test passes on x86.
>> So, the only thing I can propose now is to disable this test on PPC.
>> (Maybe it's not that bad, I was going to kill at least some of the
>> related code this year).
>>
>> --kcc
>>
>> On Tue, Jan 24, 2017 at 8:44 AM, Bill Seurer <seurer at linux.vnet.ibm.com>
>> wrote:
>>
>> This broke the powerpc sanitizers.
>>
>> updatellvm 292861
>> ninja
>> ninja check-asan
>> Testing Time: 29.65s
>>   Expected Passes    : 457
>>   Expected Failures  : 1
>>   Unsupported Tests  : 157
>>
>> updatellvm 292862
>> ninja
>> ninja check-asan
>>
>> FAIL: AddressSanitizer-powerpc64le-linux :: TestCases/Posix/coverage-direct.cc
>> (528 of 615)
>> ******************** TEST 'AddressSanitizer-powerpc64le-linux ::
>> TestCases/Posix/coverage-direct.cc' FAILED ********************
>> Script:
>> --
>> /home/seurer/llvm/build/llvm-test2/./bin/clang --driver-mode=g++
>> -fsanitize=address -mno-omit-leaf-frame-pointer -fno-omit-frame-pointer
>> -fno-optimize-sibling-calls -gline-tables-only -m64
>> -fsanitize-coverage=func -DSHARED /home/seurer/llvm/llvm-test2/
>> projects/compiler-rt/test/asan/TestCases/Posix/coverage-direct.cc
>> -shared -o /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/
>> test/asan/POWERPC64LELinuxConfig/TestCases/Posix/Output/lib$(basename
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp).dynamic.so
>> -fPIC
>> /home/seurer/llvm/build/llvm-test2/./bin/clang --driver-mode=g++
>> -fsanitize=address -mno-omit-leaf-frame-pointer -fno-omit-frame-pointer
>> -fno-optimize-sibling-calls -gline-tables-only -m64
>> -fsanitize-coverage=func /home/seurer/llvm/llvm-test2/
>> projects/compiler-rt/test/asan/TestCases/Posix/coverage-direct.cc -ldl
>> -o /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp
>> rm -rf /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct
>> mkdir -p /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/
>> test/asan/POWERPC64LELinuxConfig/TestCases/Posix/Output/
>> coverage-direct/normal
>> env ASAN_OPTIONS=coverage=1:coverage_direct=0:coverage_
>> dir=/home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct/normal:verbosity=1
>>
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/lib$(basename
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp).
>> dynamic.so
>> /usr/bin/python2.7 /home/seurer/llvm/llvm-test2/projects/compiler-rt/lib/
>> sanitizer_common/scripts/sancov.py print /home/seurer/llvm/build/llvm-
>> test2/projects/compiler-rt/test/asan/POWERPC64LELinuxConfig/
>> TestCases/Posix/Output/coverage-direct/normal/*.sancov
>> >/home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/
>> coverage-direct/normal/out.txt
>> mkdir -p /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/
>> test/asan/POWERPC64LELinuxConfig/TestCases/Posix/Output/
>> coverage-direct/direct
>> env ASAN_OPTIONS=coverage=1:coverage_direct=1:coverage_
>> dir=/home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct/direct:verbosity=1
>>
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/lib$(basename
>> /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct.cc.tmp).
>> dynamic.so
>> cd /home/seurer/llvm/build/llvm-test2/projects/compiler-rt/test/asan/
>> POWERPC64LELinuxConfig/TestCases/Posix/Output/coverage-direct/direct
>> /usr/bin/python2.7 /home/seurer/llvm/llvm-test2/projects/compiler-rt/lib/
>> sanitizer_common/scripts/sancov.py rawunpack *.sancov.raw
>> /usr/bin/python2.7 /home/seurer/llvm/llvm-test2/projects/compiler-rt/lib/
>> sanitizer_common/scripts/sancov.py print *.sancov >out.txt
>> cd ../..
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170124/66c2d0ab/attachment.html>