Hi,
I've tried to run eu-unstrip on a few ARM coredumps and found that a
majority of them contains no (or just a few) build-ids.
The interesting thing is that it only seems to happen for koji-built
applications (I've not been exploring deeply). If I try to create a
simple binary that overflows stack, the coredump seems to be sane and
eu-unstrip output contains build-ids:
$ eu-unstrip -n --core core.3149
0x10000+0x9000 5395dd8efe2a0e3d1c3cc2c44fcfb0e1eb797aea@0x10178 . - [exe]
0x4f818000+0x13f558 6d6d788a9cd7ddd06d2d5df1d385b342d98e6aee@0x4f818184
/lib/libc.so.6 /usr/lib/debug/lib/libc-2.15.so.debug libc.so.6
0x45ed8000+0x256f0 00e18824df6ca37d5dadb4595d20149c1c5d7208@0x45ed8104
/lib/libgcc_s.so.1 /usr/lib/debug/lib/libgcc_s-4.7.0-20120416.so.1.debug
libgcc_s.so.1
On the other hand, killing /bin/sleep with SIGSEGV results into a
corrupted build-id section:
$ eu-unstrip -n --core core.2987
eu-unstrip: No modules recognized in core file
Trying some more complex application results into just a few build-ids
present. This is the output from a 100M gnome-shell coredump:
$ eu-unstrip -n --core coredump
0x407a0000+0xa000 56a17098ce67817ca727b607a9fc095a3e7597ae@0x407a0124 .
/usr/lib/debug/usr/lib/gjs-1.0/langNative.so.debug [dso]
0x42d11000+0xa000 6220ebe96d624dc6b167a68089c046bc52a474d8@0x42d11124 .
- [dso]
0x46080000+0xa000 901c631ad8e41477d1fabbda0c04e733b373ba80@0x46080124 .
/usr/lib/debug/usr/lib/libgthread-2.0.so.0.3200.1.debug [dso]
0x47128000+0xa000 3a4a3386bdc70184d5491ab8f9ba3fab4380714a@0x47128124 .
/usr/lib/debug/usr/lib/libX11-xcb.so.1.0.0.debug [dso]
The actual memory dump seems to be fine, as after installing all
relevant debuginfos manually, I'm able to generate a full backtrace.
Any thoughts on this?
Thanks,
Michal
_______________________________________________
arm mailing list
arm@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/arm
