Re: gnustep-base 1.24.7 tests fail with SIGSEGV on PPA buildbot but not locally

[Riccardo Mottola]

Hi,

René J.V. Bertin wrote:
> Hi,
>
> Hope I can post this without having to subscribe to a ML ...
>
> I've been trying to create a PPA with gnustep packages built with clang-3.5, 
> blocks support and (preferably) full optimisation for my AMD Fam10 CPUs.
> I've based the gnustep-base packaging on Ubuntu's for 1.24.0, used with v 
> 1.24.7 source.
> After I discovered dpkg-build has to be called without options for parallel 
> building I finally got the process to go through, and uploaded it to launchpad. 
> I was hoping that the few tests that fail on my setup with its eclectic mix of 
> US, UK and French localisation would succeed on the buildbot.
>
> Instead, all tests crash: 
> https://launchpadlibrarian.net/191828513/buildlog_ubuntu-trusty-amd64.gnustep-base_1.24.7-1ubuntu3-rjvb-ppa20141204b_UPLOADING.txt.gz
> --- Running tests in base/Functions ---
> Running base/Functions/NSByteSwapping.m...
> /usr/bin/gnustep-tests: line 308:  8232 Segmentation fault      (core dumped) 
> $RUN_CMD
> Failed file:     NSByteSwapping.m aborted without running all tests!
>
> base/Functions/NSByteSwapping.m:
> Failed file:     NSByteSwapping.m aborted without running all tests!
> Running base/Functions/NSGeometry1.m...
> /usr/bin/gnustep-tests: line 308:  8256 Segmentation fault      (core dumped) 
> $RUN_CMD
> Failed file:     NSGeometry1.m aborted without running all tests!
>
>
> Evidently it wouldn't make much sense to call gnustep-tests with --debug on a 
> buildbot, but is there a way to do KDE DrKonqi-style post-mortem debugging just 
> to obtain a backtrace? I already ruled out the possibility of a CPU-feature 
> incompatibility by rebuilding everything (gnustep-make, libobjc2 and 
> gnustep-base) without a -march= option.
tests leave core files!

I also supect some of your configure options

./configure CC="clang" CXX="clang++" LD="clang" CPPFLAGS="-D_FORTIFY_SOURCE=2" 
CFLAGS="-g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -O3 -fvectorize -g -fblocks 
-fPIC -pthread" \
            OBJCFLAGS="-g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat 
-Werror=format-security -O3 -fvectorize -g -fblocks -fPIC -pthread" 
LDFLAGS="-Wl,-Bsymbolic-functions -Wl,-z,relro " \
            --disable-bfd --disable-mixedabi --with-installation-domain=SYSTEM

could you leave out all the CFLAGS and CPPLAGS as a test?

Riccardo