Date: Sun, 5 Apr 1998 15:49:34 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Peter Gerwinski <peter AT agnes DOT dida DOT physik DOT uni-essen DOT de>
cc: djgpp AT delorie DOT com
Subject: Re: I/O problems
In-Reply-To: <6g53vf$ess13@mx2.hrz.uni-essen.de>
Message-ID: <Pine.SUN.3.91.980405154916.5976T-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On 4 Apr 1998, Peter Gerwinski wrote:

>   * `gpc' invokes `gpc1'.  When `gpc1' now invokes `gpc' again
>     [using pexecute()] which in turn invokes `gpc1', the second
>     instance of `gpc1' sometimes reads nonsense [using getc()];
>     for example the first two or three characters of a line
>     somewhere in the middle of an input file get lost.
> 
>   * Sometimes also `gas' reports "I/O error" when invoked
>     the same way (from the second instance of `gpc') or
>     complains that `c:/temp/ccaaaaaa.s' could not be opened.

It is very hard to say something meaningful about these problems,
since they involve large programs and happen only ``sometimes''.
Perhaps if you could describe what is special about the times when the
problems do happen, it would help.

One possibility is that these problems are somehow due to conflicts
between names of temporary files used by the compiler driver.  The
current versions of library functions which create temporary files
aren't 100% safe when several programs run concurrently.  Perhaps you
run other instances of the compiler or other DJGPP programs which use
temporary files at the same time?

>   * I do not know whether this is related: `ls' sometimes
>     fails to retrieve a file that is definitely there

Once again, please try to understand what conditions cause `ls' to not
report some of the files.