[RFC] Infinite backtraces...

Joel Brobecker brobecker@adacore.com
Fri Dec 3 18:49:00 GMT 2004


> I've a finish up a patch that checks for this:
> >    #6  0x7aee0f08 in __pthread_create_system () from /usr/lib/libpthread.1
> >    #7  0x00000000 in ?? ()
> I.e., a zero pc unwound from a normal frame.  It is ``tricky'' to test 
> though :-(

That's pretty much what I have done as a proof of concept.
I ran it through the testsuite on our HP machine, and we got
3 new passes (from 3 unresolved, due to the expect buffer filling up).

> The other thing that would help here is for glibc's CFI to identify the 
> return-address (and CFA) column as unknown (assuming I've got my CFI 
> term correct) on the outer most frame.  It would then be easy for 
> dwarf2-unwind to identify this.  It's been discussed, agreed, but not 
> implemented.

This is of course a good solution, provided that you can use dwarf2.
On 32bit HP/UX, we're stuck.

> Right, but it shouldn't need an additional method.  The per-architecture 
> unwinder, when it detects a frame that the ABI specifies as final, 
> should return a null frame ID.  For instance, the PPC ABI explicitly 
> specifies that it's stack be terminated with a zero SP.

I am not sure this is doable. Is it? Let me check that again. Perhaps
it's ok to create the frame object, but then later compute a null frame
ID for it? As far as I remember, the sequence of events is like this
when trying to build the frame chain:

  . get_prev_frame (this_frame):
      . get_frame_id (this_frame)
          . frame_id (next_frame, this_cache)
      . check this frame ID
      . build previous frame
        (frame ID unset)
 
And then, after building each new frame, we display the information
for that new frame.

> Finally, a more long term suggestion is that we add a mechanism for
> creating or adding attributes to symbols (for instance for signal
> trampolines).  An atribute of such a symbol could be that it is
> outermost.

But could we determine that a symbol is outermost. And couldn't the
same symbol be used in both contexts?

-- 
Joel



More information about the Gdb-patches mailing list