Issue 31494: Valgrind suppression file (original) (raw)

Created on 2017-09-16 20:50 by Aaron Michaux, last changed 2022-04-11 14:58 by admin. This issue is now closed.

Messages (4)
msg302353 - (view)	Author: Aaron Michaux (Aaron Michaux) *	Date: 2017-09-16 20:50
The file "Misc/valgrind-python.supp" doesn't work on Linux x64, running Python3.5.2, when configured as such: --with-pydebug --with-valgrind --without-pymalloc Running the interpreter echo "x = 1; print(x)" | valgrind --tool=memcheck --leak-check=full --track-origins=yes --verbose --log-file=valgrind.log --suppressions=Misc/valgrind-python.supp /path/to/built/python3 Gives the attached valgrind.log file. The tail of this file says: ==22828== ==22828== LEAK SUMMARY: ==22828== definitely lost: 0 bytes in 0 blocks ==22828== indirectly lost: 0 bytes in 0 blocks ==22828== possibly lost: 298,529 bytes in 690 blocks ==22828== still reachable: 605,872 bytes in 4,874 blocks ==22828== suppressed: 0 bytes in 0 blocks ==22828== Reachable blocks (those to which a pointer was found) are not shown. ==22828== To see them, rerun with: --leak-check=full --show-leak-kinds=all ==22828== ==22828== ERROR SUMMARY: 256 errors from 256 contexts (suppressed: 0 from 0) ==22828== ERROR SUMMARY: 256 errors from 256 contexts (suppressed: 0 from 0)
msg302759 - (view)	Author: Terry J. Reedy (terry.reedy) *	Date: 2017-09-22 18:01
3.5 only gets security fixes, and this does not seem to be a security issue. Please retest with 3.6 or 3.7.
msg302760 - (view)	Author: Stefan Krah (skrah) *	Date: 2017-09-22 18:08
The suppressions are mainly for "invalid access" due to a gc trick. The primary goal is to be definitely-lost-clean. I would say it works as expected.
msg302761 - (view)	Author: Stefan Krah (skrah) *	Date: 2017-09-22 18:12
Also, I don't think --with-pydebug works well with valgrind. It's either: 1) --with-pydebug 2) CFLAGS="-O0 -g" --with-valgrind 3) CFLAGS="-O0 -g" --without-pymalloc Combining 2) and 3) probably does not hurt, but is not necessary.