This is the mail archive of the
elfutils-devel@sourceware.org
mailing list for the elfutils project.
Re: dwarflint vs C++
- From: Roland McGrath <roland at redhat dot com>
- To: elfutils-devel at lists dot fedorahosted dot org
- Date: Wed, 08 Jul 2009 13:28:12 -0700
- Subject: Re: dwarflint vs C++
> Right. So why is it silent? Oh, right, because the coverage checking
> code is turned off by #undef. I remember doing that, I think it was
> something with me not being able to check NOP runs (which were often
> uncovered IIRC).
I don't think those should ever be there in runs longer than sh_addralign
might require. If they are, it is probably indeed suspicious.
> I don't think I wrote checking just in the CU low_pc/high_pc range.
> I'll do that and see how it turns out.
IMHO that's the only thing that makes sense as a default check. Those are
ranges clearly emitted by a compiler using -g, so checking there is good.
Other ranges (e.g. crt1.o) are just not mentioned in any DWARF at all, and
that is normal. It's good to have a switch to enable "complain if any code
has no DWARF", which e.g. I think we can try on kernel binaries or on
glibc's own statically-linked binaries (their .debug files) like ldconfig.
(There we build all the assembly with -g too, so everything should be covered.)
Thanks,
Roland