From:	Andris Pavenis <pavenis AT latnet DOT lv>
To:	djgpp-workers AT delorie DOT com, Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
Subject:	Re: binutils-2.14
Date:	Mon, 23 Jun 2003 12:11:34 +0300
User-Agent:	KMail/1.5.2
References:	<200306181709 DOT 02598 DOT pavenis AT latnet DOT lv> <200306182039 DOT 36545 DOT pavenis AT latnet DOT lv> <3EF0EA8C DOT 54E541E AT phekda DOT freeserve DOT co DOT uk>
In-Reply-To:	<3EF0EA8C.54E541E@phekda.freeserve.co.uk>
MIME-Version:	1.0
Message-Id:	<200306231211.34342.pavenis@latnet.lv>
Reply-To:	djgpp-workers AT delorie DOT com

On Thursday 19 June 2003 01:41, Richard Dawe wrote:
> Hello.
>
> Andris Pavenis wrote:
> [snip]
>
> > Most significant DJGPP related change is adding workaround for
> > bug when seeks over file end causes garbage written to file. This problem
> > caused bootstrap comparisson error when bootstrapping current CVS
> > versions of GCC-3.4 and also once when bootstrapping GCC-3.3 release
> > (perhaps only when GCC-3.3 were also installed).
> >
> > As we don't have fix yet even in CVS version of DJGPP, then perhaps
> > such temporary workaround is needed.
>
> Hmmm, I thought it was fixed in CVS.
>
> Evidence from <libc/fd_props.h>:
>
> /* Tell write and _write to test for file offset greater than EOF.  If so,
>    they will fill the gap with zeroes.  */
> #define FILE_DESC_ZERO_FILL_EOF_GAP 0x02
>
> I'm sure this worked before. I'm sure I checked it once, when I was messing
> with the fd_props and open code.
>
> Have you built binutils 2.14 against DJGPP CVS without your patch?

I built it using last update of DJGPP v2.03.

I haven't tested current CVS, but I think that binutils linked with last  
released version (at current moment) of libc.a must work. Therefore
workaround was needed.

Andris

- Raw text -