Message-ID: <35AB115F.8778570F@catalysis.nsk.su>
Date: Tue, 14 Jul 1998 14:05:51 +0600
From: Alexey Yakovlev <jack AT catalysis DOT nsk DOT su>
Organization: Inst. of Catalysis
MIME-Version: 1.0
To: Dave Love <d DOT love AT dl DOT ac DOT uk>
CC: djgpp-workers AT delorie DOT com, fortran AT gnu DOT org
Subject: Re: Inconsistencies between g77 v0.5.23 and v0.5.19
References: <Pine DOT SUN DOT 3 DOT 91 DOT 980712192223 DOT 27114O-100000 AT is> <rzqpvfb2aq8 DOT fsf AT djlvig DOT dl DOT ac DOT uk> <35A9B2D3 DOT 63A0D300 AT catalysis DOT nsk DOT su> <rzq3ec5r38s DOT fsf AT djlvig DOT dl DOT ac DOT uk>
Content-Type: text/plain; charset=koi8-r
Content-Transfer-Encoding: 7bit
Precedence: bulk

Dave Love wrote:
> 
> >>>>> "AY" == Alexey Yakovlev <jack AT catalysis DOT nsk DOT su> writes:
> 
>  AY> Changes to f/runtime always included undefining NON_ANSI_RW_MODES
>  AY> and specifying some other flags like -DWANT_LEAD_0, -DPad_UDread
>  AY> -DALWAYS_FLUSH, etc.
> 
> Is there some particular reason to define ALWAYS_FLUSH for DJGPP?  I
> don't know why it should be any different than for Unix.

  No particular reason. Just to have a possibility to watch at the
program output. In my scientific calculations there is always need to
have the very latest information how the program runs.

>  AY> configur.bat in gcc-2.8.1 is some relic from previous versions.
>  AY> For DJGPP is simply calls 'sh djconfig.sh'.

 Sorry, but this is from Andris e-mail

 AFAIK, the script was introduced by Andris Pavenis when building
gcc-2.8.1 on DJGPP.

>  AY> 3. a new tmpfile() should be included in libg2c built for
>  AY> DJGPP. It is a quick and dirty way but it should work.
> 
> I'm not sure that's something I'd want to do.  Is there something
> about it at delorie.com or elsewhere that we could reference in the
> problems list?
  A bug report and workaround are here:
  http://www.delorie.com/djgpp/bugs/show.cgi?000143

  You don't have to do anything because most (all?) g77/DJGPP people use
precompiled binaries downloaded from Simtel.net. The patched tmpfile()
may be inserted into libg2c.a by Andris or Eli (or somebody else) after
compilation but before uploading to simtel. When DJGPP v2.02 is out it
may be removed from the library. Then you don't have to ask people to
patch their libc and recompile it. I'm sure that many people wouldn't
want to download libc sources just to recompile one function. Especially
this concerns Fortran-under-DOS programmers. Many of them even never
heard of a `patch' utility. They certainly will wait until corrected
library appears (or switch to another compiler/platform).
 It's a different problem if moving the function from one library to
another violates some rule or copyright or whatever. 
 
Regards
Alexey