[Bug gas/32435] [2.44 Regression] gas produces unaligned sections on arm-linux-gnueabihf for kernel build

[nickc at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=32435

--- Comment #26 from Nick Clifton <nickc at redhat dot com> ---
(In reply to Jan Beulich from comment #22)
> (In reply to Nick Clifton from comment #20)
> > Plus, if I have read Jan's v2 patch correctly, sections in object files will
> > still be aligned, but just aligned to the architecture's minimum file
> > alignment, rather than the sections specified alignment.  This means that
> > the sections can still be fetched using aligned reads, and so the problem of
> > a bus error resulting from an unaligned read should no longer occur.
> 
> No, there are cases where no (file) alignment would be enforced at all:
> Specifically for the one caller of
> _bfd_elf_assign_file_position_for_section() who passes "false" for "align"
> and "0" for "log_file_align".

Ah, sorry I had missed that.

> I'm pretty certain that Linux modpost is (was)
> tripped by this case. An option might by to also pass bed->s->log_file_align
> there, just that apart from the issue here there didn't appear to be a
> reason to do so.

Would passing bed->s->log_file_align instead of 0 have a large impact on object
file size ?  (My guess is that it would not unless the object file has a very
large number of sections in it).  If so, then passing bed->s->log_file_align
would be a reasonable compromise.

-- 
You are receiving this mail because:
You are on the CC list for the bug.