From: Andris Pavenis <pavenis AT latnet DOT lv>
To: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
Subject: Re: ANNOUNCE: DJGPP packages of gdb 5.3
Date: Mon, 13 Jan 2003 12:50:56 +0200
User-Agent: KMail/1.5
Cc: djgpp AT delorie DOT com
References: <200212291526 DOT gBTFQJU25879 AT delorie DOT com> <200301071120 DOT 23447 DOT pavenis AT latnet DOT lv> <3E21518F DOT 9F54AC8A AT phekda DOT freeserve DOT co DOT uk>
In-Reply-To: <3E21518F.9F54AC8A@phekda.freeserve.co.uk>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200301131250.57037.pavenis@latnet.lv>
Reply-To: djgpp AT delorie DOT com

On Sunday 12 January 2003 13:29, Richard Dawe wrote:
> Hello.
>
> Andris Pavenis wrote:
> > On Thursday 02 January 2003 13:02, Richard Dawe wrote:
> > > Hello.
> > >
> > > andris AT hal DOT astr DOT lu DOT lv wrote:
> > > [snip]
> > >
> > > > I didn't have enough time to try to dig into more details and to find
> > > > where something is wrong in debug info. I only modify GDB not to
> > > > generate internal error in this case. Of course it's ugly, but I
> > > > don't think that there is another way before GCC is fixed.
> > >
> > > Yes, a work-around seems like the best way right now.
> > >
> > > > Unfortunatelly I don't have a workaround patch here around me now. So
> > > > I can send it only after I return to Finland next week (I'm currently
> > > > in Latvia for Christmass and New Year holydays)
> > >
> > > Since gdb 5.1.1 and 5.2.1 exhibit the same problem, I decided that the
> > > port of gdb 5.3 should be released anyway. When you get back, I'd be
> > > happy to do a new release of the port with your patch. Have a good
> > >  holiday!
> >
> > This problem were not present in 5.1.1 and 5.2.1.
> >
> > I'm back in Finland, so can send patch I applied for gdb-5.3 prereleases.
> > I put #ifdef __DJGPP__ there although I reproduced the same problem also
> > under Linux when -imacros were used.
>
> [snip]
>
> I don't think we're talking about the same bug. Your patch doesn't fix the
> problem I reported with mixed-debug formats and tab completion of symbol
> names. What problem does your patch fix?

It's a problem with debug info about preprocessor defines and DWARF2 debugging 
info, so it oftem (but not in all cases) occur when -g3 (-gdwarf2-3) is 
specified:
  
preprocessor option  -imacros ... (we use it from specs to get sys/version.h 
in) causes GCC-3.X (at least GCC-3.1 qand GCC-3.2.1) missreport in debug info
include from the 1st line of source file as being from 2nd one. If the same 
symbol is defined in include file really included from line 2 (eg. with 
proper #undefine not to have warnings), then one is getting GDB internal 
error. My patch simply silences GDB, not to complain about that. Even if we 
would now have a fix for GCC, it would be best to add this workaround.

> I have a new release ready, but I want to be sure what I am releasing. It's
> a big upload over a 56kbps modem.

Andris