Time spent on elfutils: 50%
That was mostly dwarflint. I started work on arange coverage analysis
done only on CU-covered portions of address space.
To decide whether uncovered portion is acceptable alignment, I need to
reintroduce sections into the coverage analysis, because in general, CUs
can cover various sections.
Doing the above I found a couple buglets in various corner cases that
weren't exercised during my mass testing run.
I seem to be unable to keep the promised 80% on elfutils. I think those
consistently observed 50% are closer to what might be expected this
Look whether dwarf output code stabilized and see if I can find any XXXs
or unfinished business to poke in.
Finish dwarflint arange coverage analysis.
On Fedora rawhide binaries containing ifunc symbols are now marked as
ELFOSABI_LINUX. This makes elflint accept such binaries:
2009-07-13 Mark Wielaard <mjw(a)redhat.com>
* elflint.c (check_elf_header): Accept ELFOSABI_LINUX.
Fixes the run-elflint-self.sh check when run on fedora rawhide.