From: cgf AT cygnus DOT com (Christopher Faylor)
Subject: Re: winsup-990126: Strange problem with cvs 1.10
1 Feb 1999 06:23:35 -0800
Message-ID: <19990201085041.E25385.cygnus.cygwin32.developers@cygnus.com>
References: <36B25554 DOT DE26DD67 AT cityweb DOT de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
To: Corinna Vinschen <corinna DOT vinschen AT cityweb DOT de>,
        DJ Delorie <dj AT delorie DOT com>
Cc: cygwin32-developers AT cygnus DOT com

On reflection, I think that this is probably the problem with fopen that
was noted previously.  fopen defaults to using O_TEXT if no 'b' flag is
specified.

I checked in a patch to newlib for this but I don't think that it
has been reflected in a snapshot yet.  Maybe tonight.

-chris

On Sun, Jan 31, 1999 at 02:15:12AM +0100, Corinna Vinschen wrote:
>DJ Delorie wrote:
>> 
>> Does it display the same symptoms without CYGWIN=binmode?
>
>Hi DJ,
>
>yes, it does. I have locally solved this problem by replacing the
>code, which sets binary/text mode in fhandler_base::fhandler_base() and
>fhandler_base::open() by the old code from winsup-981230.
>
>Regards,
>Corinna
>
>
>> > > This sounds like it might be a CR/LF problem.  Could you pipe the
>> > > output through "cat -vetu" and see if it notices ^M changes?
>> >
>> > Yes, you can see ^M in the versionfile foo,v while output. This
>> > isn't the truth for the file foo,v on disk, but cvs copies the
>> > contents into a temporary file, which has dos newlines after that.
>> > Then it's opened for reading in binary mode!
>> >
>> > I'm using CYGWIN='binmode ntea tty' and only binary mounts.

-- 
cgf AT cygnus DOT com
http://www.cygnus.com/