Re: Strange VC/CVS failure

[Andre Spiegel]

Jason Rumney <address@hidden> writes:

> A recent change I made to gnus-art.el did not get checked in when I
> followed the usual VC checkin procedure.

Hmm, what exact "procedure" is it you are using?  Typing C-x v v,
entering a log message and typing C-c C-c, which should perform the
check-in?  Do you use CVSREAD or some such to make unmodified CVS
files read-only?

I assume this is with a remote repository?

Does VC actually ask you for a log message before the check-in fails?

What is the state of the file after the check-in has failed?

The next time it happens, could you have a look at the *vc* buffer and
see what output was produced by CVS?

A remote theory that vaguely springs to my mind is that the check-in
might fail because the repository contains more recent changes and
these need to be merged into your workfile first.  It could be that
this merge actually happens, but you are not properly informed of it.
After any recent changes have been picked up into your working file,
the check-in will finally succeed.

With the new mode vc-cvs-stay-local, which is on by default, VC no
longer queries the repository before the actual check-in (that used to
be different in Emacs 20).  Maybe there's something going wrong here.
I'll have a careful look at the code tomorrow, concerning that
possibility, but maybe these hints can already help you to chase it
down.