From: pavenis AT lanet DOT lv
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>, djgpp-workers AT delorie DOT com
Date: Thu, 9 Mar 2000 18:54:54 +0200
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: ANNOUNCE: Binutils 2.9.5 beta 1 released
Message-ID: <38C7F37E.13286.10468D7@localhost>
References: <Pine DOT A41 DOT 4 DOT 05 DOT 10003041201240 DOT 21590-100000 AT ieva05 DOT lanet DOT lv>
In-reply-to: <Pine.SUN.3.91.1000305100552.29735W-100000@is>
X-mailer: Pegasus Mail for Win32 (v3.12c)
Reply-To: djgpp-workers AT delorie DOT com

On 5 Mar 2000, at 10:08, Eli Zaretskii wrote:

> 
> On Sat, 4 Mar 2000, Andris Pavenis wrote:
> 
> > About CRC failures in bnu2951s.zip in Simtelnet:
> > 	I also experienced similar problems with some attempts to upload
> > 	gcc related archives to ftp.delorie.com (and as I remember I was
> >         not alone):
> 
> FWIW it never happened to me.  Very strange, these failures.
> 
> >    	So perhaps we should always send MD5SUMs to DJ for tests and he
> >         should try to find a weak place where problems appears.
> 
> Probably a good idea, assuming that DJ has time to verify ;-)
> 

I got such problem once with one gcc-2.95.X related archive, Mark got 
it and I remeber somebody else got it.

 - It may be problem at our end (like broken FTP client). In this case   
   testing of MD5 summs on ftp.delorie.com would help to avoid
   mirroring it. However also testing ZIP archive would be Ok

> > 3) Is it really needed to patch configure (as it is generated file). If 
> >    one applies patches to different binutils snapshot he/she is expected
> >    to know how to use autoconf. At least patch for configure failed
> >    as perhaps You used file generated by DJGPP port of autoconf  
> 
> I usually run a Sed script on configure, not Patch.  A Sed script is
> much more robust that a patch file, because it depends less on
> context, and because you control how much context do you want it to
> pick.
> 
> In any case, I think it is undesirable to ask people who want to
> build the distribution to install Autoconf.  We already ask too much
> for building packages; adding Autoconf, Automake, M4, and Perl to that
> would be unjustified (IMHO).
> 

Well I could divide all binutils users into 3 groups:
	- ordinary users, who installs binary archive
       - ones who builds binutils from bnu*s.zip
	- ones who patches binutils snapshot themselves and 
          builds it

I don't think that it too much to ask to have autoconf for 3rd group 
only. Similary if one builds gcc-2.95.2 from gcc2952s.zip autoconf is 
not needed, if one gets gcc2952s2.zip and original FSF sources then
autoconf is required.

Andris