Re: [rfa] eliminate some annoying mdebug-related symtab crashes

[Daniel Jacobowitz <dmj+ at andrew dot cmu dot edu>]

On Fri, Jul 20, 2001 at 10:24:25PM -0400, Elena Zannoni wrote:
> Oh dear, I feel ill. Where is the paper bag, quick!  So, the || check
> before calling init_psymbol_list works if the two lists are kind of in
> synch. They are grown/allocated together.  They are both zero at
> start, and they get initialized both to the same size in
> init_psymbol_lists.

Exactly.

> Mon Apr  8 23:57:43 1991  John Gilmore  (gnu at cygint.cygnus.com)
> 
>         * dbxread.c (dbx_symfile_read):  Initialize psymbol list if this
>         is the first symbol read, even if not mainline.
> 
> I would think that extend_psymbol_list was introduced afterwards, and
> this broke the logic of Gilmore's change.
> 
> So what to do?  I am kind of convinced that your patch is
> correct. But..., is there any chance you can run the gdb testsuite
> before and after your changes on a couple of platforms that have at
> least dwarf2 and stabs debug info (I mean separately, not in the same
> objfile)?

I'll try to do this and get back to you about it next week.

> Another alternative would be to insert a call to
> init_psymbol_lists(objfile, NUM_OF_SYMBOLS) somewhere in mdebugread.c.

I didn't do that because I couldn't find a reasonable NUM_OF_SYMBOLS,
but I could just use a hardcoded guess like some of the other symbol
readers do, I suppose.  I'd rather fix it as I proposed; I'll run those
testsuites and see what I can turn up.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer