-
Bug
-
Resolution: Unresolved
-
Major
-
None
-
rhel-9.6, rhel-9.7
-
None
-
No
-
Important
-
rhel-plumbers
-
ssg_core_services
-
5
-
False
-
False
-
-
None
-
Red Hat Enterprise Linux
-
None
-
None
-
None
-
Unspecified
-
Unspecified
-
Unspecified
-
-
All
-
None
This is reopened bugzilla Bug 1942064 "libmagic is not thread-safe when ~/.magic is used"
see https://bugzilla.redhat.com/show_bug.cgi?id=1942064
The issue is still present, see e.g.
https://artifacts.osci.redhat.com/testing-farm/df7a1b9e-1807-4131-a3bf-9bc973565827/,
namely
https://artifacts.osci.redhat.com/testing-farm/df7a1b9e-1807-4131-a3bf-9bc973565827/work-CI-Tier-1x7xqhxkw/plans/internal/CI-Tier-1/execute/data/guest/default-0/Sanity/rpmbuild-dependency-generators-sanity-test-56/output.txt
Processing files: dummy-pkg-1.0-1.noarch unknown, 0: Warning: using regular magic file `/root/.magic' unknown, 0: Warning: using regular magic file `�p��' unknown, 0: Warning: using regular magic file `/root/.magic' error: magic_load failed: could not find any valid magic files! unknown, 0: Warning: using regular magic file `/root/.magic' error: Recognition of file "/root/rpmbuild/BUILDROOT/dummy-pkg-1.0-1.x86_64/usr/bin/dummy_execpattern_fileB.sh" failed: mode 100755 could not find any valid magic files! warning: Recognition of file "/root/rpmbuild/BUILDROOT/dummy-pkg-1.0-1.x86_64/usr/lib64/dummy-pkg/dummy_path_fileB" failed: mode 100644 could not find any valid magic files!
It's not easy to reproduce the issue. It sometimes occurs when running this test:
http://pkgs.devel.redhat.com/cgit/tests/rpm/tree/Sanity/rpmbuild-dependency-generators-sanity-test
In the bugzilla Comment 4, there's a hint for a fix provided by Panu Matilainen:
> get_default_magic(void) > { > static const char hmagic[] = "/.magic/magic.mgc"; > static char *default_magic; > char *home, *hmagicpath; > > #ifndef WIN32 > struct stat st; > > if (default_magic) { > free(default_magic); > default_magic = NULL; > }
That's not going to work in a multi-threaded program. default_magic would have to be moved to the magic_open()'ed context, but that might not be entirely trivial as it's exposed via libmagic API in a way that doesn't include the context.
- causes
-
RHEL-90851 rpm test optimizations to support glibc preload script
-
- Closed
-
