Date: Sun, 19 Jul 1998 13:16:13 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Jos De Laender <De_Laender-De_Winter AT village DOT uunet DOT be>
cc: djgpp AT delorie DOT com
Subject: Re: GNAT - tasking - FP emulation
In-Reply-To: <35AE557B.4BFC@village.uunet.be>
Message-ID: <Pine.SUN.3.91.980719131537.32A-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Thu, 16 Jul 1998, Jos De Laender wrote:

> - GNAT307 (GNU ADA) on a 386DX system without 387. ( some people
> 	do like old-timers, don't they ?)
> - GNU WMEMU387 installed as SW emulator after problems with
> the djgpp EMU387 (FPE if I remember correctly)

What were those problems, exactly?

In general, I would suggest to stick with EMU387 if at all possible,
since WMEMU387 is AFAIK not maintained by anybody, and thus could
become incompatible as DJGPP evolves.

> >From the moment I use tasking in my ADA (i.e. the diners - example)
> I crash with :
> 	cannot continue from exception; exiting due to signal 0123 
> 	Page fault at ...
> 	(stuff)

Did you try to avoid using FP code in the program?  If so, did it
help?

Since both multi-tasking and FP emulator use exceptions, you could
have some kind of conflict between them.  The error message you cite
means that an exception handler for a real exception (as opposed to
ones DJGPP fakes to generate SIGINT) tried to return to the place
where the exception originated.  This isn't allowed, so the DJGPP
exception-handling machinery aborts your program.

> Now, in a brave mood, I wanted to dig into code and try to debug.
> However, from the moment I use a debugger (gdb or fsdb) and just
> run the ada-tasking problems, everything goes OK !! i.e. the 
> diners-example runs under debugger and crashes standalone.
> 
> Here I'm stuck ... Anyone who can assist ?

Try the good old printf debugging.