Re: Checking new mail very slow with current CVS

[Simon Josefsson <jas@extundo.com>]

Russ Allbery <rra@stanford.edu> writes:

>> Is 5.8 seconds slow?
>
> Yes, it's roughly five times longer than it took under 5.8.7 as an
> unmeasured estimate.  The slowdown was really dramatic; it's not just a
> matter of a half-second here and there.

Then probably the marks stuff isn't to blame.  Hm.  For comparison,
could you get the similar elp results from Gnus 5.8.7?  (Be sure to
not read lots of mail using it, those marks will be lost when you
revert to Oort Gnus.  Maybe I should fix this.)

>> I guess the new marks stuff makes up for 1 second of that time.  Hm.
>> But `nnml-request-update-info' takes 0.5 seconds longer than
>> `nnml-marks-changed-p' which should only happen when Gnus starts the
>> first time (or if some other Gnus changed the files).  Then the marks
>> stuff should only be bound by the extra time spent in
>> `nnml-marks-changed-p'.  You could try adding a 'nnml-marks-is-evil'
>> server parameter.
>
> I'm not quite following what you're saying here; do you think this is due
> to marks?  Setting nnml-marks-is-evil and then pressing "g" again resulted
> in no noticeable time difference, but is that too late to change the
> setting?

It should make it 1s (or .5s) faster judging from your elp results. If
you instrument it, you should see that `nnml-request-update-info' and
`nnml-marks-changed-p' doesn't take any measurable time, if
`nnml-marks-is-evil' is non-nil.