Hi!

On Fri, 2 Mar 2012 18:18:10 -0700, Kevin Buettner <kevinb@redhat.com> wrote:
> On Fri, 02 Mar 2012 12:17:39 +0100
> Thomas Schwinge <thomas@codesourcery.com> wrote:
> > (I'm assuming ``this'' is the 28 instructions limit.)  This is what many
> > other architectures' tdep files are doing, too, and it triggers only as a
> > safeguard if we fail to find any other limit.  Also, a similar thing is
> > used in sh_in_function_epilogue_p.
> 
> In the past, [...]

Thanks for telling the history of that.

> enhancements which remove those limits would be welcome.

Should we file a bug in bugzilla for that?


> > > We do need some limit though.  I'm just concerned about debugging leaf
> > > functions where that limit will put us into the next function.  (This
> > > was one of the problems with my earlier patch - it didn't handle that
> > > case.)
> > 
> > As I said, this limit is only a safeguard if everything else fails.
> > Before that, the end of the function will have tried to be determined
> > with the symbol table (find_pc_partial_function), or debug information
> > (skip_prologue_using_sal), which will typically trigger (but not in PLT
> > slots).
> 
> Hmm.  Have you tried it when only minimal symbols are present?  (It is
> useful to have this stuff working so that you can get decent stack
> traces when only linker symbols are present.)

Yes, I had a look at gdb.base's arrayidx, advance and nodebug examples,
and they're working as before.


Grüße,
 Thomas