Re: [RFA] Don't SEGV on invalid dwarf2 frame info

[Elena Zannoni <ezannoni at redhat dot com>]

Mark Kettenis writes:
 > Richard Henderson <rth@twiddle.net> writes:
 > 
 > > Elena, this is the patch I was thinking about.
 > > 
 > > For the audience, there is at least one bug in current cvs ld's .eh_frame
 > > optimization code that results in padding being added between sections.
 > > But we saw similar problems when we added support for .eh_frame generation
 > > within the assembler (and didn't .align sections), so the discussion in 
 > > the patch is a bit more broad than that.
 > > 
 > > Does this seem reasonable?
 > 
 > It does to me.  It's unfortunately that this is necessary, but it
 > makes things more robust, so please check this in.
 > 
 > Mark

As is, the patch doesn't compile. I'll have updated one in a minute. I
also think that there should be a check after reading the
initial_length, that such length is within the section. In my case,
that's where gdb encounters the first bogus value.

elena