test_faulthandler hangs at 100% CPU, gave it 18 hours before cancelling. 6642 pts/2 S+ 0:00 /var/tmp/portage/dev-lang/python-3.3.5-r1/work/x86_64-pc-linux-gnu/python -bb -R -E -Wdefault -Werror::BytesWarning -m test.regrtest --slaveargs [["test_faulthandler", 0, false], {"debug": false, "use_resources": [], "failfast": false, "huntrleaks": false, "match_tests": null, "timeout": null, "output_on_failure": false}] 10535 pts/2 R+ 18:03 /var/tmp/portage/dev-lang/python-3.3.5-r1/work/x86_64-pc-linux-gnu/python -E -c import faulthandler faulthandler.enable() faulthandler._stack_overflow() I see this new warning during the build: x86_64-pc-linux-gnu-gcc -pthread -Wno-unused-result -DNDEBUG -O2 -march=native -pipe -fwrapv -I. -IInclude -I/var/tmp/portage/dev-lang/python-3.3.5-r1/work/Python-3.3.5/Include -fPIC -DPy_BUILD_CORE -c /var/tmp/portage/dev-lang/python-3.3.5-r1/work/Python-3.3.5/Modules/faulthandler.c -o Modules/faulthandler.o ../Python-3.3.5/Modules/faulthandler.c: In function 'stack_overflow': cc1: warning: function may return address of local variable [-Wreturn-local-addr] ../Python-3.3.5/Modules/faulthandler.c:903:19: note: declared here unsigned char buffer[4096]; which led me to this commit: https://hg.python.org/cpython/rev/689092296ad3 Bug: http://bugs.python.org/issue23433
Note as the bug mentions this will also manifest at runtime as python segfaults.
"... fixed ..." upstream : http://bugs.python.org/issue23433#msg235745
(In reply to Ryan Hill from comment #1) this code only seems to apply to tests. the fix is in the python-3.5.0 release and the 3.4 branch (i guess they didn't bother putting it in the 3.3 branch). i've added the 3.4 patch to the 3.4/3.3 ebuilds: http://gitweb.gentoo.org/repo/gentoo.git/commit/?id=bd3586765e47208668f27f707ec7f6d06b5d92a8
