Re: Problem building CVS-1.10.x

["Steven N. Hirsch" <shirsch at adelphia dot net>]

On Tue, 28 Dec 1999, Earnie Boyd wrote:

> > I updated to most recent cygwin1-YYYYetc.. and cywin-inst.  No change in
> > (mis)behavior.  It would be great if someone could double-check me on
> > this by downloading any recent cvs newer than 1.10.5 or so and building
> > with the configure defaults.  If you can check modules from a unix-based
> > CVS server, you're further than I'm getting.
> > 
> 
> Great.  No, I've not used cvs in this manner.  How is the UNIX based
> server being accessed.  What it sounds like you're running into is the
> M$ version of a line ending, i.e.: \r\n.  This can be avoided if you
> ensure that the access to the server is in binary mode processing.  
> Set CYGWIN="... tty binmode ..." before starting bash.  Use the binary
> mode mounts for all mounts. Caveat to this, all text files must have
> UNIX style endings.
> 

<sigh>, no my immediate problem is _not_ the line endings.  It's the fact
that I'm unable to compile a working version of cvs-1.10.x.  It builds
without incident or error.  When I try to use it by doing, e.g.

$ cvs co CVSROOT

(against my own server)

it spits out:

cvs.exe [checkout aborted]: writing to server: The descriptor is a file,
 not a socket

and dies.  This is the problem I need to solve.  Once I can get it working
to begin with, I'll worry about modifying it to preserve line-endings.

For whatever reason, no one else is reporting this runtime error.  The
error is coming from a section of code that tries to flush a buffer back
to the server in client.c of cvs.

Steve



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com