[PATCH] Fix setting-breakpoints regression on PPC64 (function descriptors)

Yao Qi qiyaoltc@gmail.com
Fri Dec 8 16:57:00 GMT 2017


On Fri, Dec 8, 2017 at 11:34 AM, Pedro Alves <palves@redhat.com> wrote:
> That's fine with me.  I guess we end up with the wrong section
> in the function descriptor / PPC64 case (".opd" instead of some kind
> of  ".text" where the resolved function lives), but it shouldn't
> matter, since the old code did that as well, AFAICT.

I tested the patch on gcc110, there is no regression.  I pushed it in.

>
> (I noticed that get_sal_arch doesn't consider sal.objfile, probably
> because it predates addition of the 'obfile' field.  We could probably
> fill in / use that field more, but we don't need to do that now.)

I noticed that too, but one thing I am not sure is that sal.objfile is
*only* used for probe,

  /* The probe associated with this symtab_and_line.  */
  probe *prob = NULL;
  /* If PROBE is not NULL, then this is the objfile in which the probe
     originated.  */
  struct objfile *objfile = NULL;

so can we use it in other cases (when probe is not used)?  I don't
know.  If so, are sal.objfile and sal.section->objfile different or same?
I need to look at the code there.

-- 
Yao (齐尧)



More information about the Gdb-patches mailing list