>>>> "Stephen" == Stephen J Turnbull
Stephen> Raymond Toy writes:
> Not sure what's going on. I restarted xemacs and tried
> I get an error in the smtp trace buffer that I need to start STARTTLS.
> But starttls-use-gnutls is nil.
Stephen> This is weird. Same error as before?
Stephen> BTW, ISTR that there was a report of issues with smtpmail/gnutls on
Stephen> emacs-devel. They've recently started linking to the underlying
Stephen> library instead of using a process, maybe it was something to do with
Stephen> that switchover, but maybe it was associated with the same
Stephen> start-process method we use.
> Various experiments to investigate don't help, but I finally
> decided to just byte-compile smtpmail.el again. And then suddenly
> a fresh xemacs will automatically start starttls. Hurray!
Stephen> Maybe the byte compiler (more likely some macro?) is evaluating
Stephen> starttls-use-gnutls and stashing the value in the byte-compiled file.
> I still get the error about being unable to delete the process
> the end of smtpmail-via-smtp. I stuck a (message "Deleteing process")
> and (sit-for 1) just before (delete-process process). I don't have
> this problem anymore. Some kind of race condition?
Stephen> That's quite possibly the reason for byte-compiled code failing where
Stephen> interpreted code doesn't. The race condition might be aggravated by
Stephen> multicore systems, too, if smtpmail isn't properly waiting for the
Stephen> subprocess to terminate, but assumes that it will do so "fast
What is the correct xemacs way to wait for a process to terminate?
XEmacs-Beta mailing list