This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] On bulletproofing -fsyntax-only
- To: Geoff Keating <geoffk at cygnus dot com>
- Subject: Re: [PATCH] On bulletproofing -fsyntax-only
- From: Jeffrey A Law <law at cygnus dot com>
- Date: Mon, 29 May 2000 11:07:25 -0600
- cc: Nix <nix at esperi dot demon dot co dot uk>, gcc-patches at gcc dot gnu dot org
- Reply-To: law at cygnus dot com
In message <jmhfbiv5w0.fsf@envy.cygnus.com>you write:
> > I think that the method currently being used to suppress output within
> > gcc (preventing asm_out_file from being written to when syntax_only is
> > on if it looks like that code path will be walked down) is fragile,
> > because unless *every* use of asm_out_file is armoured that way, we'll
> > eventually end up trying to write to it again. And armouring every use
> > is tricky and ugly, because ports manipulate asm_out_file, and because
> > not every use of asm_out_file is via the ASM_*() macros.
>
> I agree...
Likewise.
> > I propose bulletproofing the existing code by making asm_out_file point
> > to /dev/null (or a comparable bitbucket) when syntax_only is on. With
> > that in place, we'll waste nothing but a little time if we accidentally
> > try to stick stuff into the asm_out_file, rather than crashing.
>
> This sounds like a reasonable idea.
We can't assume anything about /dev/null. Consider windows, vms and other
non-unixy systems. Ahh, y'all seem to have thought of it below. Excellent.
> > I foresee only two potential problems with this approach: different OSes
> > have different bit buckets (some might not have them at all), and you
> > can't normally seek in a bit bucket.
>
> On systems that really don't have a bit bucket, you could write to a
> temporary file, and just delete it when done.
That would be fine by me. Presumably we'd write an autoconf test to see
if we've got a bit bucket.
> > Here's a patch which squashes this problem using the method above
> > (without the /dev/null autoconf test, as yet). If needed I can mail off
> > a copyright assignment form (although, especially without the autoconf
> > part, this fix is probably too small to merit it). I've also got a patch
> > against 2.95.2, because my original problem was there ;)
>
> Please do send in a form.
Agreed.
jeff