glibc-2.37.9000-13.fc39 introduced an annocheck regression when compared to -10.fc39: MAYB: test: lto, reason: a region of code compiled without LTO was detected(abi-note.c) FAIL: fortify test because -D_FORTIFY_SOURCE=[2|3] was not present on the command line (abi-note.c) Can be seen in the CI tests (fedora-ci.koji-build.rpminspect.static-analysis of https://bodhi.fedoraproject.org/updates/FEDORA-2023-3317068e8d). Reproducible: Always
This is probably yet another exception to add for glibc sources; glibc is not built with _FORTIFY_SOURCE.
(In reply to Václav Kadlčík from comment #0) > glibc-2.37.9000-13.fc39 introduced an annocheck regression when > compared to -10.fc39: > > MAYB: test: lto, reason: a region of code compiled without LTO was > detected(abi-note.c) > FAIL: fortify test because -D_FORTIFY_SOURCE=[2|3] was not present on the > command line (abi-note.c) > > Can be seen in the CI tests (fedora-ci.koji-build.rpminspect.static-analysis > of https://bodhi.fedoraproject.org/updates/FEDORA-2023-3317068e8d). Which architecture(s) are affected? All of them?
(In reply to Siddhesh Poyarekar from comment #1) > This is probably yet another exception to add for glibc sources; glibc is > not built with _FORTIFY_SOURCE. We use this kludge in glibc.spec: %if 0%{?_annotated_build} > 0 # libc_nonshared.a cannot be built with the default hardening flags # because the glibc build system is incompatible with # -D_FORTIFY_SOURCE. The object files need to be marked as to be # skipped in annobin annotations. (The -specs= variant of activating # annobin does not work here because of flag ordering issues.) # See <https://bugzilla.redhat.com/show_bug.cgi?id=1668822>. BuildFlagsNonshared="-fplugin=annobin -fplugin-arg-annobin-disable -Wa,--generate-missing-build-notes=yes" %endif We could probably build with -D_FORTIFY_SOURCE=3 instead, given that this is for the non-shared bits only (where fortification has no impact).
Hmm, I suppose we could even make this wait for a bit (i.e. until we enable fortification in 2.38) and then actually fix this by enabling fortification. I think the first cut excludes csu, but that can be done incrementally.
(In reply to Florian Weimer from comment #2) > Which architecture(s) are affected? All of them? Yes
It looks like some more additions are needed to annocheck's heuristics for detecting glibc binaries. The "/dev/null" component name in some of annocheck's messages is coming from the new string based notes which record a filename based upon the gcc command line used to build the binary being checked. Presumably glibc has some kind of clever method for building some of its files.
Should be fixed with annobin-12.13-1.fc39
Yes, all the files annocheck complained about: /lib64/ld64.so.2 /lib64/ld-linux-x86-64.so.2 /lib64/libc.so.6 /lib/ld64.so.1 /lib/ld-linux-aarch64.so.1 /lib/ld-linux.so.2 /lib/libc.so.6 /usr/lib64/crt1.o /usr/lib64/gcrt1.o /usr/lib/crt1.o /usr/lib/gcrt1.o pass cleanly with annobin-12.13-1.fc39.
This bug appears to have been reported against 'rawhide' during the Fedora Linux 39 development cycle. Changing version to 39.