MIME-Version: 1.0
In-Reply-To: <201207130044.q6D0iXDf015563@envy.delorie.com>
References: <CAA-ihx9exryc-z2z30W3E+WsJG_rERTKjrMbE9D6=WbAvyrf-Q AT mail DOT gmail DOT com>
	<201207130044 DOT q6D0iXDf015563 AT envy DOT delorie DOT com>
Date: Fri, 13 Jul 2012 00:15:56 -0500
Message-ID: <CAA-ihx9yh+mLmuZqpMOn9zaG2VVvt=dc=d8oNNWbyzr-NmdvuA@mail.gmail.com>
Subject: Re: mktemp() fails on "detmp.xxx" but not "detmp.XXX" (proposed patch)
From: Rugxulo <rugxulo AT gmail DOT com>
To: djgpp-workers AT delorie DOT com
Content-Type: text/plain; charset=ISO-8859-1
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
Precedence: bulk

Hi,

On Thu, Jul 12, 2012 at 7:44 PM, DJ Delorie <dj AT delorie DOT com> wrote:
>
> CONFORMING TO
>        4.3BSD, POSIX.1-2001.  POSIX.1-2008 removes the specification
>        of mktemp().

Interesting, typical, funny, annoying that 2008 removes it, heh.
Actually, Emacs' callproc.c uses mkstemp(), which for DJGPP actually
calls mktemp() itself.

> My POSIX spec is older than 2001 so I can't confirm this.  Note that
> the XXXXXX part is *not* a filename component, but a pattern to be
> replaced, so the "files are case ignoring" argument doesn't apply.

I'm not understanding you. Or maybe you misunderstood me, dunno. Yes,
I know it's a template, and I've recompiled mktemp.c with -DTEST
several ways. My point is that mktemp() fails when Emacs
"accidentally" (in SFN 8.3 situations only) lowercase'd the template,
hence "detmp.xxx" [sic] fails where "detmp.XXX" would succeed.

Since DOS' file system (FAT) is case insensitive anyways, it seems
silly (to me) for mktemp() to fail this call just because the 'x'
chars aren't uppercase. Lemme get on my other computer for a clearer
example:

(runs to other room)

[ FreeDOS ] G:\TONY>gcc -x c -s -O -DTEST -DTOUPPER=toupper mktemp.new -o b
[ FreeDOS ] G:\TONY>gcc -x c -s -O -DTEST mktemp.new -o a
[ FreeDOS ] G:\TONY>a
/usr/foo/detmp.XXX -> /usr/foo/deaaaaaa
/usr/foo/detmp.xxx ->
[ FreeDOS ] G:\TONY>b
/usr/foo/detmp.XXX -> /usr/foo/deaaaaaa
/usr/foo/detmp.xxx -> /usr/foo/debaaaaa
[ FreeDOS ] G:\TONY>scrndump c:\tmp\emacs\mktemp.bug

See what I mean? It rejects the template with lowercase 'x'. I know
POSIX is more concerned with case sensitivity than we are, but I don't
think it's any advantage to be this picky on FAT (DOS, SFN 8.3). It
just makes hidden bugs harder to find. And it's an easy fix, but I
guess it's your call.