PCL-CVS:- strange messages in *cvs* buffer

PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

With one repository that I am using, I am getting a lot of strange
messages coming out in my *cvs* buffer. 


For example....


In directory describer/src/uk/ac:
Message: 
Message: 
Message: 
Message: 
In directory describer/src/uk/ac/man/cs/img/osclient/data:
Message: 
Message: File: Workflow.java    	Status: Up-to-date
Message: 
Message:    Working revision:	1.1.1.1
Message:    Repository revision:	1.1.1.1	/usr/local/cvs/mygrid/mygrid/describer/src/uk/ac/man/cs/img/osclient/data/Workflow.java,v
Message:    Sticky Tag:		(none)
Message:    Sticky Date:		(none)
Message:    Sticky Options:	(none)
Message: 
Message:    Existing Tags:
Message: 	rtag                     	(revision: 1.1.1.1)
Message: 	vtag                     	(branch: 1.1.1)
Message: 
              Up-To-Date  1.1.1.1
              describer/src/uk/ac/man/cs/img/osclient/data/package.html


or

Message: 
Message: 
Message: 	No Tags Exist
Message: 


and so on. Often there are enough of these that its makes it hard to
see what is going on. The results seem to be different everytime,
which makes it hard to work out what is happening. 

Has anyone got any ideas what might be the problem?

Cheers

Phil

Re: PCL-CVS:- strange messages in *cvs* buffer

[Kai Großjohann]

Phillip Lord <p.lord@russet.org.uk> writes:

> With one repository that I am using, I am getting a lot of strange
> messages coming out in my *cvs* buffer. 

What does `cvs update -dP' print in that directory?  Maybe there the
messages are easier to see and it's easier to see what's going on.

-- 
~/.signature is: umop ap!sdn    (Frank Nobis)

Re: PCL-CVS:- strange messages in *cvs* buffer

[Stefan Monnier <foo@acm.com>]

>>>>> "Phillip" == Phillip Lord <p.lord@russet.org.uk> writes:
> With one repository that I am using, I am getting a lot of strange
> messages coming out in my *cvs* buffer.

I believe this is another manifestation of a bad interaction between
glibc, ssh, and cvs.

Is it on a system that uses glibc ?
Are you accessing the repository via SSH ?


        Stefan

Re: PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

>>>>> "Stefan" == "Stefan Monnier <foo@acm.com>" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu> writes:

>>>>> "Phillip" == Phillip Lord <p.lord@russet.org.uk> writes:
  >> With one repository that I am using, I am getting a lot of
  >> strange messages coming out in my *cvs* buffer.

  Stefan> I believe this is another manifestation of a bad interaction
  Stefan> between glibc, ssh, and cvs.

  Stefan> Is it on a system that uses glibc ?  Are you accessing the
  Stefan> repository via SSH ?

Yes, and yes. Both client and server are running on redhat (7.2
client, 8 server). Is there anything that can be done to fix this? 

Phil

Re: PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

>>>>> "Kai" == Kai Gro_johann <kai.grossjohann@uni-duisburg.de> writes:

  Kai> Phillip Lord <p.lord@russet.org.uk> writes:

  >> With one repository that I am using, I am getting a lot of
  >> strange messages coming out in my *cvs* buffer.

  Kai> What does `cvs update -dP' print in that directory?  Maybe
  Kai> there the messages are easier to see and it's easier to see
  Kai> what's going on.


I've tried running cvs from the command line, and examining the
output. As far as I can tell, it's all fine. The caveat here is that
the errors are different every time, so I don't know which bit of the
output to examine by hand, and there is a lot of it. 

Phil

Re: PCL-CVS:- strange messages in *cvs* buffer

[Stefan Monnier <foo@acm.com>]

>>> With one repository that I am using, I am getting a lot of
>>> strange messages coming out in my *cvs* buffer.
Stefan> I believe this is another manifestation of a bad interaction
Stefan> between glibc, ssh, and cvs.
Stefan> Is it on a system that uses glibc ?  Are you accessing the
Stefan> repository via SSH ?
> Yes, and yes. Both client and server are running on redhat (7.2
> client, 8 server). Is there anything that can be done to fix this? 

Complain to all three groups so that hopefully some fix will
show up at some point.   In the mean time, the Emacs code on the trunk
has the (process-connection-type nil) line changed to the following:

	       ;; If process-connection-type is nil and the repository
	       ;; is accessed via SSH, a bad interaction between libc,
	       ;; CVS and SSH can lead to garbled output.
	       ;; It might be a glibc-specific problem.
	       ;; Until the problem is cleared, we'll use a pty rather than
	       ;; a pipe.
	       ;; (process-connection-type nil) ; Use a pipe, not a pty.


-- Stefan


PS: yes, the problem only happens if the output is a pipe rather than a pty.
    it's all due to incompatible buffering settings.

Re: PCL-CVS:- strange messages in *cvs* buffer

[Kai Großjohann]

Phillip Lord <p.lord@russet.org.uk> writes:

> I've tried running cvs from the command line, and examining the
> output. As far as I can tell, it's all fine. The caveat here is that
> the errors are different every time, so I don't know which bit of the
> output to examine by hand, and there is a lot of it. 

Why is there a lot of output?  It should just print a list of dir
names, if everything is up to date.  Do you have hundreds of
directories there?  If so, why not try `cvs update -dP' in a subdir
of your complete working directory?

-- 
~/.signature is: umop ap!sdn    (Frank Nobis)

Re: PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

>>>>> "Stefan" == "Stefan Monnier <foo@acm.com>" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu> writes:

  >>>> With one repository that I am using, I am getting a lot of
  >>>> strange messages coming out in my *cvs* buffer.
  Stefan> I believe this is another manifestation of a bad interaction
  Stefan> between glibc, ssh, and cvs.  Is it on a system that uses
  Stefan> glibc ?  Are you accessing the repository via SSH ?
  >> Yes, and yes. Both client and server are running on redhat (7.2
  >> client, 8 server). Is there anything that can be done to fix
  >> this?

  Stefan> Complain to all three groups so that hopefully some fix will
  Stefan> show up at some point.  In the mean time, the Emacs code on
  Stefan> the trunk has the (process-connection-type nil) line changed
  Stefan> to the following:

  Stefan> 	       ;; If process-connection-type is nil and the
  Stefan> 	       ;; repository is accessed via SSH, a bad
  Stefan> 	       ;; interaction between libc, CVS and SSH can
  Stefan> 	       ;; lead to garbled output.  It might be a
  Stefan> 	       ;; glibc-specific problem.  Until the problem
  Stefan> 	       ;; is cleared, we'll use a pty rather
  Stefan> than
  Stefan> 	       ;; a pipe.  (process-connection-type nil) ; Use
  Stefan> 	       ;; a pipe, not a
  Stefan> pty.

Ta. I shall try this bit out, and see if it solves my woes. 

Phil

Re: PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

>>>>> "Kai" == Kai Gro_johann <kai.grossjohann@uni-duisburg.de> writes:

  Kai> Phillip Lord <p.lord@russet.org.uk> writes:

  >> I've tried running cvs from the command line, and examining the
  >> output. As far as I can tell, it's all fine. The caveat here is
  >> that the errors are different every time, so I don't know which
  >> bit of the output to examine by hand, and there is a lot of it.

  Kai> Why is there a lot of output?  It should just print a list of
  Kai> dir names, if everything is up to date.  Do you have hundreds
  Kai> of directories there?  If so, why not try `cvs update -dP' in a
  Kai> subdir of your complete working directory?


Sorry, I meant from a cvs status. I tried cvs update -dP, and it all
appears to work fine. 

If, as Stefan suggested, its a problem with piping, then it would work
fine on tty? 

Cheers

Phil

Re: PCL-CVS:- strange messages in *cvs* buffer

[Phillip Lord]

>>>>> "Stefan" == "Stefan Monnier <foo@acm.com>" <monnier+gnu.emacs.help/news/@flint.cs.yale.edu> writes:


  Stefan> 	       ;; If process-connection-type is nil and the
  Stefan> 	       ;; repository is accessed via SSH, a bad
  Stefan> 	       ;; interaction between libc, CVS and SSH can
  Stefan> 	       ;; lead to garbled output.  It might be a
  Stefan> 	       ;; glibc-specific problem.  Until the problem
  Stefan> 	       ;; is cleared, we'll use a pty rather
  Stefan> than
  Stefan> 	       ;; a pipe.  (process-connection-type nil) ; Use
  Stefan> 	       ;; a pipe, not a
  Stefan> pty.



  Stefan> PS: yes, the problem only happens if the output is a pipe
  Stefan> rather than a pty.
  Stefan>     it's all due to incompatible buffering settings.



Okay, so I've tried this, and it does seem to eliminate the
problem. Thanks very much for your help!

Cheers

Phil