Mailing-List: contact new-httpd-help@apache.org; run by ezmlm
Precedence: bulk
Reply-To: new-httpd@apache.org
Message-ID: <022101bfbf42$e72e9640$c1e01b09@raleigh.ibm.com>
From: "Bill Stoddard" <stoddard@raleigh.ibm.com>
To: <new-httpd@apache.org>
References: <000201bfbf39$607f9bc0$345985d0@corecomm.net>
Subject: Re: ap_os_is_filename_valid calls.
Date: Tue, 16 May 2000 10:27:45 -0400
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit

We cannot afford the overhead of an additional stat. Also consider that mod_file_cache
does a trick to avoid doing the stat if the file is cached.

Bill
----- Original Message -----
From: William A. Rowe, Jr. <wrowe@lnd.com>
To: <new-httpd@apache.org>
Sent: Tuesday, May 16, 2000 9:18 AM
Subject: RE: ap_os_is_filename_valid calls.


> I've looked at this and ment to fix... if someone beats me to it
> though;
>
> We can drop all the device driver name test garbage.  The file
> attribute value 0x80 flags the file as a 'virtual', or something
> that we never want apache to test.  Ergo... walking the list of
> directories and names, if we stat with an 0x80, we are in the
> wrong place and need to fail.  NUL, CON, LPTx etc all return this
> value, since the ancient days of MS-DOS.
>
> Bill
>
> > -----Original Message-----
> > From: Bill Stoddard [mailto:stoddard@raleigh.ibm.com]
> > Sent: Tuesday, May 16, 2000 7:23 AM
> > To: new-httpd@apache.org
> > Subject: Re: ap_os_is_filename_valid calls.
> >
> >
> >
> > > On Mon, May 15, 2000 at 08:06:47AM -0400, Bill Stoddard wrote:
> > > > I think canonical_filename is very application specific
> > and does not
> > > > belong in APR.
> > >
> > > Hmmm. Actually, I can see it being generally useful.
> > >
> > > Let's take an app on the complete opposite side of the map from
> > > Apache: say, a word processor. It needs to know the last 4 files
> > > edited so it can present a history. Now, wouldn't it be
> > nice if those
> > > 4 files were actually distinct? demonical_filename can be useful in
> > > this case, too.
> > >
> > > Any app that needs to check if two filenames are intended
> > to point to
> > > the same file can use this (or a variant of what we have,
> > anyway). So
> > > it seems that APR would be a good place for it.
> >
> > Have you looked at what all ap_os_case_canonical_filename
> > does in src/os/win32/util.c?  I
> > don't think the word processor would need to spend cycles
> > checking for all these cases.  I
> > don't have strong thoughts on it one way or the other, just
> > seems we have bigger fish to
> > fry (like getting the buildconf, libtool, etc. fixed).
> >
> > Bill
> >