Red Hat Bugzilla – Bug 90837
emacs-21.3-3 is having weird problems with process sentinels
Last modified: 2015-01-07 19:05:01 EST
I have an emacs-lisp function which calls an external process in a comint
buffer and installs a sentinel to finish up the task when the external process
exits (it's a spam parser and complaint generator).
I recently upgraded from emacs-21.3-1 to emacs-21.3-3. After doing so, the
sentinel sometimes exhibits new, broken behavior -- when the process is
finished and the sentinel is called, (current-buffer) returns the rmail buffer
I was in before calling the spam parser when it SHOULD return the comint buffer
in which the parser was running. I did not change anything in my code to cause
this new behavior -- it's definitely something that changed either between
emacs-21.3-1 and emacs-21.3-3 or (less likely) in glibc or something else that
Furthermore, sometimes after the spam parser exits and my sentinel restores my
previous window configuration, the next character I type in the rmail buffer
isn't interpreted as an rmail command as it should be -- instead I get a beep
and a message in the minibuffer telling me that the buffer is read-only. If I
hit the same key again, the second time it's interpreted correctly as an Rmail
command. M-x view-lossage doesn't show anything that might explain this
I've tried to create a test case of this problem but I've been unable to do
so. I feel a little silly filing a bug report about a weird problem like this
with no associated test case, but I wanted to give you a heads up that there
may be something in emacs-21.3-3 that "isn't quite right."
What locale are you using? (Most of the changes made between -1 and -3
relate to utf-8 support.)
Can you reproduce the problem in a vanilla Emacs session (ie
"emacs -q --no-site-file")?
I have LANG=en_US and LC_COLLATE=C. Both of these problems manifest themselves
when I run emacs -q --no-site-file and then load my .emacs file manually (which
I need to do to get the spam-bouncing functions defined there).
I don't think we have any process related patches -
so I suggested taking this upstream.
Have you tried with cvs emacs?
Wait. The only major change between -1 and -3 was the addition
of MuleUCS for utf-8 support...
What happens when you run emacs with --no-site-file?
I no longer use the emacs-lisp functions that exhibit this behavior,
and I don't have time to resurrect them to attempt to duplicate this,
so I'm going to close this with WORKSFORME. Perhaps it'll happen to
someone else and they'll be able to come up with a good test case, in
which case they can file a new bug or reopen this one.