The problem seems to be with sysdig. The name mangling on the symbol shows that
sysdig is expecting a function that takes two bools ('...Ebb'), but libsinsp.so
only provides a function that takes one bool ('...Eb'). On questing we can see
the mismatch:
$ readelf -Ws /usr/bin/sysdig | grep _ZN5sinsp16set_import_users
51: 0000000000000000 0 FUNC GLOBAL DEFAULT UND
_ZN5sinsp16set_import_usersEbb
$ readelf -Ws /usr/lib/x86_64-linux-gnu/libsinsp.so.0 | grep
_ZN5sinsp16set_import_users
7274: 0000000000323420 15 FUNC GLOBAL DEFAULT 14
_ZN5sinsp16set_import_usersEb
This single bool version matches the actual upstream source of libsinsp
here:
https://github.com/falcosecurity/libs/blob/27082051530e9e67c4ce291b6b3c7ddb842c5d10/userspace/libsinsp/sinsp.cpp#L426,
so is presumably the correct one.
On noble, sysdig is looking for the single bool version (which matches the lib)
and the problem isn't present:
$ readelf -Ws /usr/bin/sysdig | grep _ZN5sinsp16set_import_users
60: 0000000000000000 0 FUNC GLOBAL DEFAULT UND
_ZN5sinsp16set_import_usersEb
$ readelf -Ws /usr/lib/x86_64-linux-gnu/libsinsp.so.0 | grep
_ZN5sinsp16set_import_users
7274: 0000000000323420 15 FUNC GLOBAL DEFAULT 14
_ZN5sinsp16set_import_usersEb
I also checked on resolute and this problem isn't present - and again sysdig is
looking for the correct single bool version here:
$ readelf -Ws ./sysdig | grep set_import
72: 0000000000000000 0 FUNC GLOBAL DEFAULT UND
_ZN5sinsp16set_import_usersEb
sysdig was bumped from 0.38.1+repack to 0.40.0+repack between questing
and resolute, so the fix is in there somewhere.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2110564
Title:
P/Q: ubuntu_sysdig_smoke_test 3 failures
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-kernel-tests/+bug/2110564/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
