Re: nnimap process die repeatedly

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Juan José García-Ripoll <juanjose.garciaripoll@gmail.com> writes:

> Hi,
>
> I have set up nnimap successfully and was using it fine with 26.3. I've
> since upgraded to 27.1 (last RC) and I see nnimap process dying
> often. Sometimes it is not noticed, other times gnus complains and I
> have to close and reopen the server. In both cases a new buffer with a
> new number appended to it is created. As of today, 6 hours into my
> working day, I have 26 buffers with the name " *nnimap" in it, all for
> the same server. Is there any way to debug why this happens? Why are
> buffers not reused or processes deleted when this happens?

I'll look in to cleaning up the process buffers, but would you be
willing to try something? Visit the nnimap.el source file, find the
definition of `nnimap-open-connection-1' (probably line 439), go a
couple lines down, and change this bit:

(run-at-time (* 60 15) (* 60 15)

to:

(run-at-time (* 60 3) (* 60 3)

(or otherwise something smaller) Then re-evaluate the function and run
it for a day? nnimap processes ought to have TCP keepalive enabled, and
also should be sending "NOOP" IMAP commands to the server on the timer
specified by the number above. A shorter timer might solve your problem,
in which case it might be useful to make this number user-configurable.

But regardless, those dead process buffers shouldn't be left lying
around.

Eric