[compiler-rt] r244136 - [sanitizer] Add the flag handle_sigfpe that is default true to handle SIGFPE crashes same as SIGSEV crashes, patch by Karl Skomski

Thu Aug 6 10:54:11 PDT 2015

Oh, the lit test is also affected:

/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/stage1/./bin/clang
-gline-tables-only -fsanitize=address -mcpu=cortex-a15 -mthumb
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/llvm/projects/compiler-rt/test/sanitizer_common/TestCases/Linux/fpe.cc
-o /home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/stage1/projects/compiler-rt/test/sanitizer_common/asan-arm-Linux/Linux/Output/fpe.cc.tmp
not          /home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/stage1/projects/compiler-rt/test/sanitizer_common/asan-arm-Linux/Linux/Output/fpe.cc.tmp
2>&1 | FileCheck --check-prefix=CHECK1
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/llvm/projects/compiler-rt/test/sanitizer_common/TestCases/Linux/fpe.cc
ASAN_OPTIONS=handle_sigfpe=0 not --crash
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/stage1/projects/compiler-rt/test/sanitizer_common/asan-arm-Linux/Linux/Output/fpe.cc.tmp
2>&1 | FileCheck --check-prefix=CHECK0
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/llvm/projects/compiler-rt/test/sanitizer_common/TestCases/Linux/fpe.cc
ASAN_OPTIONS=handle_sigfpe=1 not
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/stage1/projects/compiler-rt/test/sanitizer_common/asan-arm-Linux/Linux/Output/fpe.cc.tmp
2>&1 | FileCheck --check-prefix=CHECK1
/home/linaro/devel/buildbot/clang-cmake-thumbv7-a15-full-sh/llvm/projects/compiler-rt/test/sanitizer_common/TestCases/Linux/fpe.cc
--
Exit Code: 2

Command Output (stderr):
--
FileCheck error: '-' is empty.

Is sigfpe generated under different circumstances on ARM?
r244234 re-lands this patch w/o the redundant unit test bit
(AddressSanitizer-Unit::AddressSanitizer.FPECrashTest)
and with "REQUIRES: x86_64-supported-target"
in test/sanitizer_common/TestCases/Linux/fpe.cc

I am online for the rest of the day (PST) and watching the bots.

On Thu, Aug 6, 2015 at 10:27 AM, Kostya Serebryany <kcc at google.com> wrote:

> +Karl, author of the patch.
>
> Thanks for reverting and sorry for the breakage.
> I don't have access to ARM myself, nor much experience with it.
> Any ideas what might have gone wrong?
>
> I am tempted to re-apply the patch w/o the AddressSanitizer.FPECrashTest
> test -- the same functionality
> is also covered by a lit test which did not seem to fail on ARM. WDYT?
>
>
> On Thu, Aug 6, 2015 at 5:43 AM, Renato Golin <renato.golin at linaro.org>
> wrote:
>
>> Sorry, reverted in r244210, we should investigate what's wrong offline.
>>
>> cheers,
>> --renato
>>
>> On 6 August 2015 at 10:04, Renato Golin <renato.golin at linaro.org> wrote:
>> > On 5 August 2015 at 22:19, Kostya Serebryany <kcc at google.com> wrote:
>> >> Author: kcc
>> >> Date: Wed Aug  5 16:19:11 2015
>> >> New Revision: 244136
>> >>
>> >> URL: http://llvm.org/viewvc/llvm-project?rev=244136&view=rev
>> >> Log:
>> >> [sanitizer] Add the flag handle_sigfpe that is default true to handle
>> SIGFPE crashes same as SIGSEV crashes, patch by Karl Skomski
>> >
>> > Hi Kostya,
>> >
>> > It seems this commit broke our ARM bots:
>> >
>> >
>> http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-full/builds/6917
>> >
>> >
>> http://lab.llvm.org:8011/builders/clang-cmake-thumbv7-a15-full-sh/builds/1885
>> >
>> > You haven't been warned because there were other failures in there,
>> > that are now gone.
>> >
>> > cheers,
>> > --renato
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150806/28d07b95/attachment.html>