Re: Lost process output in pipe between Emacs and CVS

[Derek Robert Price]

Stefan Monnier wrote:

> > > The problem is that people use cvs 2>&1 with CVS_RSH set to ssh, and
> > > expect it to work, and think it is a CVS bug when it fails.  Educated
> > > users don't have a problem.
> > >      
> > Perhaps a quick fix then is to detect that stdio has been set to
> > nonblocking and output an error message?
> >    
>
> The circumstances under which the problem appears are sufficiently
> particular that there are tens of quick fixes available.  The question
> is how to fix it for real such that neither SSH nor CVS nor PCL-CVS
> (in Emacs) need to work around the problem at a functionality or
> performance cost.
>
> How do non-glibc systems deal with it ? (I've only seen the
> problem reported on glibc systems until now)
> It seems legitimate to use stdio on stderr without having to worry about
> some application that you have forked and that thus shared your stderr,
> so I think the problem is that glibc's stdio does not correctly handle
> the case where the file is in non-blocking mode.
>
>
>         Stefan
>  

I agree, I would prefer to see this fixed in stdio, but Paul Eggert 
objected to this in an offshoot to this thread on at least the 
bug-cvs@gnu.org mail list based on the fact that it conflicted with the 
POSIX definition: 
<http://www.mail-archive.com/bug-cvs%40gnu.org/msg04461.html>.

Paul suggests an extension to stdio to avoid conflicts with the POSIX 
spec.  Is this reasonable?

Derek

--
               *8^)

Email: derek@ximbiot.com

Get CVS support at http://ximbiot.com
--
I know of no safe depository of the ultimate powers of the society but the
people themselves, and if we think them not enlightened enough to exercise that
control with a wholesome discretion, the remedy is not to take it from them,
but to inform their discretion.

			- Thomas Jefferson, 1820.