Re: [request] please test new pcl-cvs
Sorry, I have no way to test NT at this point. Wow, I didn't
know
pcl-cvs used to work on it. That's encouraging at least...
I was surprised as well, since subprocesses were pretty broken on b38.
First, which pcl-cvs package version is broken? 1.30?
1.30 and 1.12 packages both fail.
Can you try to determine whether the problem is pcl-cvs or XEmacs by
trying the newer pcl-cvs on the older XEmacs and vice-versa?
Alas I have nuked the old install, so can't test on the old XEmacs.
However, it is noticable that all other subprocess things work better on
this XEmacs than the old one.
Which version of CVS appears first on your path (run "cvs
-v")?
1.9.26 which is the version I expect to be using.
change I made was to just use the one found on your path, and I
nuked
the kludgy code that was trying to locate CVS.
Yes, I had to patch round this on the previous versions since it was
very broken.
The message below indicates CVS is being invoked incorrectly, so I
guess it is most likely an XEmacs or choice of shell problem.
I agree that it is an invocation problem. However, it doesn't work with
cmd or bash, which are the only shells I have available (and one more
than most NT users will have!) It appears that the necessity for
redirection and quoting issues are mucking things up.
Is there a good reason for using redirection rather than processing
everything together? From a cursory look at the code it would appear
possible to send the output of the cvs command to a temporary buffer,
then use the process sentinel to parse the entire buffer into the
cookie-collection. It appears this would both make the code simpler and
remove the need to redirect, while only modifying run-cvs-process and
the process sentinels. If you think this is feasible I'll try to hack
this up.
Cheers,
Malcolm