X-Authentication-Warning:	delorie.com: mailnull set sender to djgpp-workers-bounces using -f
Date:	Thu, 27 Dec 2001 18:11:07 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender:	eliz AT is
To:	Tim Van Holder <tim DOT van DOT holder AT pandora DOT be>
cc:	djgpp-workers AT delorie DOT com, "'Charles Sandmann'" <sandmann AT clio DOT rice DOT edu>
Subject:	Re: lfn from scratch...
In-Reply-To:	<000101c18eee$b36ed3b0$1c7d76d5@zastaixp>
Message-ID:	<Pine.SUN.3.91.1011227180545.22136E-100000@is>
MIME-Version:	1.0
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

On Thu, 27 Dec 2001, Tim Van Holder wrote:

> > That won't help, I think: our chdir doesn't test that bit 
> > (and does not call stat).
> 
> Should be easy enough to add an access(X_OK) to chdir, and modify
> access() and stat() not to set that attribute for UNCs.

More slow-down...

> > The equivalent of "test -d" (IIRC, access(D_OK)) also doesn't check
> > that.  These and other cases look at the directory attribute bit,
> > not at the execute bit.
> 
> That test is irrelevant; it's the execute bit, not the directory bit
> that matters.

That's what I meant, but backwards: our library functions test the 
directory bit, so you cannot change their behavior by simulating the 
execute bit.

> > Application code will still not expect such calamities.
> 
> Then that application code is broken; we generally deal with Unix
> ports, and such permission settings are fairly common on Unices.

I don't think I ever saw a program which looks at the execute bit of a 
directory before it chdir's there.  They all test the directory bit, and 
sometimes the write bit.

> The trick there would be to know when to release the mapping.
> For cmd.exe that's easy: as soon as you 'popd' away from the UNC.
> For us, that won't be easy to determine.  Keeping it mapped until
> program end is undesirable too, especially since there are only
> 25 mappings available at best.

That's the tricky decision, yes.  We could try the LRU approach, once all 
the available letters are taken, for example.  A program which uses that 
many different UNC directories at once should be a very rare case, I think.

- Raw text -