valgrind errors in rawhide
martinpitt opened this issue · comments
The valgrind checks recently started to fail, for two different reasons. One can be reproduced in a DEBUG=1 test container with
G_DEBUG=fatal-warnings,fatal-criticals,gc-friendly MALLOC_CHECK_=3 LD_LIBRARY_PATH=. PATH=.:$PATH GI_TYPELIB_PATH=. TOP_SRCDIR=/source /source/src/umockdev-wrapper valgrind -q --leak-check=full --show-possibly-lost=no --errors-for-leak-kinds=definite --error-exitcode=33 --trace-children=yes --trace-children-skip=/bin*,/usr/bin/*,/usr/local/bin/* ./test-umockdev -p /umockdev-testbed/add_device_errors
which fails with
==2116== Conditional jump or move depends on uninitialised value(s)
==2116== at 0x4974595: g_match_info_next (gregex.c:1051)
==2116== by 0x49747AA: g_regex_match_full (gregex.c:2162)
==2116== by 0x49748D0: g_regex_match (gregex.c:2069)
==2116== by 0x48B5239: umockdev_testbed_record_parse_line (umockdev.vala:1479)
==2116== by 0x48B0318: umockdev_testbed_add_dev_from_string (umockdev.vala:1303)
==2116== by 0x48A6CF5: umockdev_testbed_add_from_string (umockdev.vala:727)
==2116== by 0x408785: t_testbed_add_device_errors (test-umockdev.c:476)
==2116== by 0x498AB2D: UnknownInlinedFun (gtestutils.c:2933)
==2116== by 0x498AB2D: g_test_run_suite_internal (gtestutils.c:3021)
==2116== by 0x498A894: g_test_run_suite_internal (gtestutils.c:3038)
==2116== by 0x498AFFA: g_test_run_suite (gtestutils.c:3112)
==2116== by 0x498359C: UnknownInlinedFun (gtestutils.c:2234)
==2116== by 0x498359C: g_test_run (gtestutils.c:2221)
==2116== by 0x41BCC9: main (test-umockdev.c:2410)
==2116==
==2116== Conditional jump or move depends on uninitialised value(s)
==2116== at 0x4974595: g_match_info_next (gregex.c:1051)
==2116== by 0x49747AA: g_regex_match_full (gregex.c:2162)
==2116== by 0x49748D0: g_regex_match (gregex.c:2069)
==2116== by 0x48B5353: umockdev_testbed_record_parse_line (umockdev.vala:1482)
==2116== by 0x48B0318: umockdev_testbed_add_dev_from_string (umockdev.vala:1303)
==2116== by 0x48A6CF5: umockdev_testbed_add_from_string (umockdev.vala:727)
==2116== by 0x408785: t_testbed_add_device_errors (test-umockdev.c:476)
==2116== by 0x498AB2D: UnknownInlinedFun (gtestutils.c:2933)
==2116== by 0x498AB2D: g_test_run_suite_internal (gtestutils.c:3021)
==2116== by 0x498A894: g_test_run_suite_internal (gtestutils.c:3038)
==2116== by 0x498AFFA: g_test_run_suite (gtestutils.c:3112)
==2116== by 0x498359C: UnknownInlinedFun (gtestutils.c:2234)
==2116== by 0x498359C: g_test_run (gtestutils.c:2221)
==2116== by 0x41BCC9: main (test-umockdev.c:2410)
==2116==
# DEBUG: umockdev.vala:1363: creating device /devices/target (subsystem foo)
==2116== Conditional jump or move depends on uninitialised value(s)
==2116== at 0x4974595: g_match_info_next (gregex.c:1051)
==2116== by 0x49747AA: g_regex_match_full (gregex.c:2162)
==2116== by 0x49760CF: g_regex_replace_eval (gregex.c:3354)
==2116== by 0x497651A: g_regex_replace_literal (gregex.c:3255)
==2116== by 0x48A2848: string_replace (glib-2.0.vapi:1548)
==2116== by 0x48A43B1: umockdev_testbed_add_devicev_no_uevent (umockdev.vala:486)
==2116== by 0x48B2FB2: umockdev_testbed_add_dev_from_string (umockdev.vala:1364)
==2116== by 0x48A6CF5: umockdev_testbed_add_from_string (umockdev.vala:727)
==2116== by 0x408785: t_testbed_add_device_errors (test-umockdev.c:476)
==2116== by 0x498AB2D: UnknownInlinedFun (gtestutils.c:2933)
==2116== by 0x498AB2D: g_test_run_suite_internal (gtestutils.c:3021)
==2116== by 0x498A894: g_test_run_suite_internal (gtestutils.c:3038)
==2116== by 0x498AFFA: g_test_run_suite (gtestutils.c:3112)
Most tests fail due to this. This smells like a bug in glib, as it happens through other code paths as well, e.g.
==3350== Conditional jump or move depends on uninitialised value(s)
==3350== at 0x4974595: g_match_info_next (gregex.c:1051)
==3350== by 0x49747AA: g_regex_match_full (gregex.c:2162)
==3350== by 0x49748D0: g_regex_match (gregex.c:2069)
==3350== by 0x48ABE3C: umockdev_testbed_load_evemu_events (umockdev.vala:1086)
Wasn't GRegex recently ported to a newer version of the PCRE library?
i.e. https://gitlab.gnome.org/GNOME/glib/-/merge_requests/2529
This is an actual issue, tracked in https://gitlab.gnome.org/GNOME/glib/-/issues/2741 . It was fixed 5 days ago, and just needs a new release now. I'll give this a few more days then, as it's awkward to ignore.
This should be fixed in glib 2.74.0, which is in rawhide, Debian unstable, and Ubuntu devel series now. And indeed the latest run is all green again! 💚