ASAN build failure with GCC 11 #36480
Comments
|
assign core |
|
New categories assigned: core @Dr15Jones,@smuzaffar,@makortel you have been requested to review this Pull request/Issue and eventually sign? Thanks |
|
A new Issue was created by @makortel Matti Kortelainen. @Dr15Jones, @perrotta, @dpiparo, @makortel, @smuzaffar, @qliphy can you please review it and eventually sign/assign? Thanks. cms-bot commands are listed here |
|
Worked around in #36493 |
|
@smuzaffar Do you think it would be worth to investigate deeper than #36493? (i.e. keep this issue open after the PR has been merged) |
|
Looks like a compiler bug. The only cmssw/FWCore/Framework/src/PathsAndConsumesOfModules.cc Lines 687 to 690 in 2438e09 generate a call If I change that to the error goes away. The second one happens at cmssw/L1Trigger/CSCTriggerPrimitives/src/CSCCathodeLCTProcessor.cc Lines 210 to 213 in 2438e09 and can be worked around by changing that to: so it seems to be a very picky bug. I'll try making a reproducible test case. Should I check in the workarounds? |
|
Thanks @dan131riley!
I'm not sure. On one hand these workarounds would allow keeping the ASAN compilation flags, on the other hand I'm slightly concerned these workarounds would quickly sediment. Maybe they would be better than adjusting compiler flags. @smuzaffar, what do you think? If done, for the |
|
@makortel I'd vote for your workaround by changing the ASAN flags for those modules (I didn't see your PR until after I made my comment), and I'll work on a bug report. |
|
@dan131riley workaround is better than adjusting the flags for these special files. @dan131riley go ahead and open a PR. meanwhile I am testing latest gcc 11.2 changes ( hopeing that they might have fixed this) |
|
Issue remains there even with newer GCC 11 [a] |
|
looks like the issue is just not with these two files. Once you fix these two files then you get a lot more e.g. see the PR tests results looks like |
|
Found this issue google/sanitizers#1324 that said (over a year ago)
|
|
I propose to drop this flag for now ( see https://github.com/cms-sw/cmssw-config/pull/88/files and cms-sw/cmsdist#7516 ) so that we can continue with ASAN builds |
|
I was thinking the same, so I agree. |
|
The build failure was worked around, so I think this issue can be closed now. |
|
+core |
|
This issue is fully signed and ready to be closed. |
cmsbuild
added
core-approved
fully-signed
and removed
core-pending
pending-signatures
labels
Otherwise compiler may produce a code that linker fails to link. For openmw it fails with following error (https://gitlab.com/elsid/openmw/-/jobs/3571943608): Linking CXX executable ../../openmw_test_suite mold: error: undefined symbol: CMakeFiles/openmw_test_suite.dir/lua/test_configuration.cpp.o: std::__cxx11::basic_stringbuf<char, std::char_traits<char>, std::allocator<char> >::_M_high_mark() const Related issue in a different project with links to GCC bug tracker: cms-sw/cmssw#36480
It seems that ASAN IBs fail to build with GCC 11 (whereas the last GCC 10 build compiled fine). The errors are in the shared object link
https://cmssdt.cern.ch/SDT/cgi-bin/buildlogs/slc7_amd64_gcc11/CMSSW_12_3_ASAN_X_2021-12-10-2300/FWCore/Framework
and
https://cmssdt.cern.ch/SDT/cgi-bin/buildlogs/slc7_amd64_gcc11/CMSSW_12_3_ASAN_X_2021-12-10-2300/L1Trigger/CSCTriggerPrimitives
The text was updated successfully, but these errors were encountered: