DWARF-2 + discarded link-once sections (was Re: Internal error in GDB)

[Daniel Jacobowitz <drow at mvista dot com>]

On Wed, Nov 20, 2002 at 11:25:26PM -0500, Jim Blandy wrote:
> Daniel Jacobowitz <drow@mvista.com> writes:
> 
> > On Thu, Nov 14, 2002 at 10:15:28AM -0500, Daniel Jacobowitz wrote:
> > > On Thu, Nov 14, 2002 at 05:34:05PM +1000, Steven Johnson wrote:
> > > > Im getting the error:
> > > > 
> > > > Internal error: pc 0x0 in read in psymtab, but not in symtab.
> > > > 
> > > > from a very recent CVS build of GDB (As of 11th November).
> > > > 
> > > > It happens when I set breakpoints in an assembler file, that has been 
> > > > assembled through GCC, using the C Preprocessor (.S file, not .s file).
> > > > 
> > > > It does nothing wrong, except pop up all the time and hence interfere 
> > > > with debugging significantly. (Im using insight and have to keep 
> > > > pressing OK to continue).
> > > > 
> > > > I searched the code and it seems to be coming from 
> > > > symtab.c:find_pc_sect_symtab (last if in the function)
> > 
> > Jim, I need your opinion on this... here's the problem.
> > 
> > 2092      /* When using the GNU linker, .gnu.linkonce. sections are used to
> > 2093         eliminate duplicate copies of functions and vtables and such.
> > 2094         The linker will arbitrarily choose one and discard the others.
> > 2095         The AT_*_pc values for such functions refer to local labels in
> > 2096         these sections.  If the section from that file was discarded, the
> > 2097         labels are not in the output, so the relocs get a value of 0.
> > 2098         If this is a discarded function, mark the pc bounds as invalid,
> > 2099         so that GDB will ignore it.  */
> > 2100      if (low == 0 && (bfd_get_file_flags (objfile->obfd) & HAS_RELOC) == 0)
> > (top-gdb)
> > 2101        return 0;
> > 2102
> > 
> > 
> > In Steven's asm is a .vectors section which goes over the PPC hardware
> > interrupt vectors in RAM - that's actually at VMA 0.  Oops.
> > 
> > Since LD still doesn't edit DWARF-2 sections when discarding, and won't
> > for the forseeable future (although it needs to eventually), we can't
> > just drop the check.  Any ideas on what to do?
> 
> Jeez.  You'd think we would have learned by now that people like to
> put things at the top and bottom of the address space...
> 
> Would it work to check that both the high and low bounds are zero?
> Obviously, that's not going to handle a zero-length section at address
> zero, but it'd at least be better than what we have now.

Nope.  The missing section base will be considered to be zero.  Thus
the size of the discarded section will be whatever it normally would
have been, at 0 and 0+size.  Isn't it gross?

> Is there *no* other way to tell whether the debug info refers to a
> discarded section?  Can we add logic to BFD to tell us about this?

There really isn't.  It would be nice to discard it at link time.  But
this is, well, really complicated.  In my inexpert opinion at least.
(This is a similar problem to coalesced symbols in Darwin, as I
understand it?)

Hmm, maybe it isn't as bad as all that for DWARF-2.  If we get a die
with low_pc and high_pc attributes referring to a discarded section, we
could discard the die and all its children.  But there might be
_references_ to that die somewhere... this requires general purpose
DWARF-2 editing that we don't really have yet.

I'm CC'ing this over to binutils, because Alan and/or Jakub would have
a much better feel for whether this is possible than I do.

> Could we generate custom GNU Dwarf 2 information so as to allow GDB to
> detect a removed section?  That is, if we *ask* the linker to tell us,
> will it?  We could make all debug info in linkonce sections children
> of new DW_TAG_GNU_linkonce dies, with magic attributes that let us
> detect whether they've been discarded.

This could work too but it would be nice to avoid it.  We were
discussing .debug_info.gnu.linkonce.* sections at one point; did we
abandon that idea because there were too many of them, or was there
some other problem?

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer