Mailing-List: contact new-httpd-help@apache.org; run by ezmlm
Message-ID: <F78ED191D292D311B5C30000C0D77400137B47@mokey.sequent.com>
From: "Turner, Ian (iant)" <iant@sequent.com>
To: "'new-httpd@apache.org'" <new-httpd@apache.org>
Subject: RE: Minor fix to v1.3.12-dev
Date: Mon, 21 Feb 2000 10:18:46 -0800
MIME-Version: 1.0
Content-Type: text/plain


   The Unix spec.'s I have seen do not call out a type/mode "t" for file
access, but do have "b" (but most implementations ignore
or noop it). If specific platforms require "t", then it should be those
platforms that specifically handle the needed item.

Bill's item 2 is probably the better way (since there are other areas this
could come into play and I believe util_script.c already
has a specific OS/2 #ifdef for this), rather than having a lot of platform
specific #ifdef's. OS/2 and Windows can either define the item on the
configuration command line or in the platform specific areas in the header
file(s).

> -----Original Message-----
> From:	William A. Rowe, Jr. [SMTP:wrowe@lnd.com]
> Sent:	Saturday, February 19, 2000 12:18 AM
> To:	new-httpd@apache.org
> Subject:	RE: Minor fix to v1.3.12-dev
> 
> This is almost an ASCII/EBCDIC issue to those cr/lf platforms, so here
> are...
> 
> TWO solutions - you pick...
> 
> One; FILE_TEXT_ARG
> 
> [global header]
> #if defined(WIN32) || defined(OS2) [|| ...]
> #define FILE_TEXT_ARG "t"
> #define FILE_RAW_ARG "b"
> #else
> #define FILE_TEXT_ARG
> #define FILE_RAW_ARG
> #endif
> 
> then all text file open calls...
> 
> [local implementation]
> hfl = fopen(fpath, "r" FILE_TEXT_ARG)
> 
> or on raw file open calls
> 
> [local implementation]
> hfl = fopen(fpath, "r" FILE_RAW_ARG)
> 
> Of course... string concat probably doesn't work on some odd ANSI compiler
> (correct???)
> 
> So solution 2:
> 
> [global header]
> #if defined(WIN32) || defined(OS2) [|| ...]
> #define FILE_TEXT_RAW_ARG
> #endif
> 
> [local implementation]
> #ifdef FILE_TEXT_RAW_ARG
> hfl = fopen(fpath, "rt")
> #else
> hfl = fopen(fpath, "r")
> #endif
> 
> 
> Thoughts?
> 
> Bill
> 
> 
> > -----Original Message-----
> > From: Lee Tibbert [mailto:Lee.Tibbert@Compaq.Com]
> > Sent: Friday, February 18, 2000 10:28 PM
> > To: new-httpd@apache.org
> > Subject: Re: Minor fix to v1.3.12-dev
> >
> >
> >
> > The non-Ansi 't' breaks the pending OpenVMS port also.  It
> > would be nice if
> > a non-standard change were #ifdef'd for the operating system which
> > used it, instead of forcing strict checking OSes to #ifdef
> > around cruft.
> >
> > We are trying to have the minimal number of #ifdef VMS entries.
> > We can #ifdef around this one too, but it lacks elegance.
> >
> > Lee
> >
> >
> > At 08:34 PM 2/18/2000 -0600, you wrote:
> > >On Mon, Feb 14, 2000 at 04:16:08PM -0800, Turner, Ian iant" wrote:
> > > >
> > > >    A minor change was made to v1.3.11 to resolve a
> > problem with Apache on
> > > > OS/2, which should have been #IFDEF'ed just for
> > > > OS/2 (and any other platform which needs the "t" mode to
> > fopen() ),
> > > since not
> > > > all implementations of Unix ignore the "t" mode.
> > > > (I.E. the change breaks Apache on some platforms)
> > >
> > >Which platforms? If OS/2 is the only platform with the Ctrl-Z issue
> > >that spawned this change (probably not true; there is Windows) then
> > >we'd use an ifdef for OS/2. If there's only one Unix that croaks on
> > >the "t", then we should ifndef for that.
> >