Date:	Fri, 24 Apr 1998 20:28:26 +0200 (WET)
From:	"Andris Pavenis (LU Astro.)" <pavenis AT lanet DOT lv>
To:	DJ Delorie <dj AT delorie DOT com>
cc:	djgpp-workers AT delorie DOT com
Subject:	Re: Generation of temporary files for gcc-2.8.1
In-Reply-To:	<199804241706.NAA27305@delorie.com>
Message-ID:	<Pine.A32.3.91.980424201812.79182A-100000@ieva05.lanet.lv>
MIME-Version:	1.0

On Fri, 24 Apr 1998, DJ Delorie wrote:

> 
> > The question is whether this approach could be acceptable for release.
> 
> Including custom replacements for libc is acceptable for any gnu
> package, since it's GPL anyway and the sources have to be available.
> 

Maybe I should be more accurate: I took mktemp.c from djlsr201.zip as base
and severaly "broke" it (from point of view of normal user).

The problem is that to fool gcc.c and choose_temp.c I had to modify
mktemp so that it is broken for any other thing (Of course I'm not putting
it back in libc.a). Therefore I'm saying the solution is ugly.
It may also cause problems with future versions if gcc.exe will call
mktemp more than once (currently it is Ok if MKTEMP_EACH_FILE is not 
defined when compiling gcc.c, I included protection against it) 
                     
The question was whether such partial solution could be used for
stable port of gcc. Anyway we are getting (I think) safe protection
from temporary filenames conflicts between copies of gcc running
simultanously. 

Andris

- Raw text -