Anyone seen anything like the following before?

extracting debug info from
/var/tmp/lazarus-0.9.24-3.fc10-root-mockbuild/usr/lib64/lazarus/lazarus
/usr/lib/rpm/debugedit:
/var/tmp/lazarus-0.9.24-3.fc10-root-mockbuild/usr/lib64/lazarus/lazarus:
Wrong directory table index 3

The last bit repeats many times; somethings the '3' is '102' instead.

This package (https://bugzilla.redhat.com/show_bug.cgi?id=187243) is
written in pascal and built using the fpc compiler. It seems that fpc
generates some kind of debug symbols, but they're not quite in the
format that the debuginfo extraction bits understand. Is this worth a
bug report, and if so, is the bug in rpm or fpc?

- J<

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

fpc generates DWARF, the same format that gcc generates and that the tools
know how to process. It might be that fpc has bugs and generates bad
DWARF. Or it might be that it generates valid DWARF but with patterns that
GCC never generates. Since most everything has DWARF generated by the same
codebase in GCC, the DWARF-consuming tools like debugedit might have bugs
that were never noticed until fpc came along.

I will try to reproduce the failing build and look at the DWARF data.


Thanks,
Roland

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list

It turns out fpc is generating invalid DWARF in the .debug_line section.
It looks to me like the data might be mostly fine, but the unit length in
the header is too short. Try {eu-,}readelf --debug-dump=line on the
unstripped binary and see it complain about the bad format.


Thanks,
Roland

--
fedora-devel-list mailing list
fedora-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/fedora-devel-list