Date: Tue, 6 Oct 1998 14:18:22 +0200 (MET DST)
From: Olivier Perron <perron AT art DOT alcatel DOT fr>
X-Sender: perron AT rtbsci146s
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Cc: M DOT A DOT Bukin AT ino DOT nsk DOT nu, nate AT cartsys DOT com, djgpp AT delorie DOT com
Subject: Re: /dev/null problems with bash when rebuild from sources
In-Reply-To: <Pine.SUN.3.91.981006143021.17118U-100000@is>
Message-Id: <Pine.GSO.4.05.9810061334420.5502-100000@rtbsci146s>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp AT delorie DOT com

On Tue, 6 Oct 1998, Eli Zaretskii wrote:

> 
> But if you set SYSROOT=C:, then it *should* work (it does for me with
> the stock bash.exe).
Yes I agree. It works with the stock bash.exe, but not with the one I've
just rebuild.

>  Please run Bash under a debugger and see what
> function fails for /dev/null.  (A good beginning would be to set a
> breakpoint in a function that reports the error, and then dsiplay the
> stack to see where did the error originate.)
> 
Here are some results: because of the value of SYSROOT being "C:",
dosutil_expand_path expands /dev/null to c:/dev/null.
Then a call to 'open' is performed with "c:/dev/null" as first argument.
It is this call which fails.

A little test program easily shows that 'stat' and 'open' effectively fail
with "c:/dev/null" but not with "/dev/null".

As I do not have the patched sources of open and stat, I cannot make
further investigations. Nate, could you please send them to me, or at
least tell me where to find them ?
 
Olivier.