Closed Bug 39810 Opened 24 years ago Closed 21 years ago

Slow shutdown

Categories

(MailNews Core :: Backend, defect, P3)

Product:
MailNews Core
Component:
Backend
Type:
defect

Tracking

(Not tracked)

Status:
VERIFIED DUPLICATE of bug 39742

People

(Reporter: BenB, Assigned: naving)

Details

(Keywords: perf) 

Linux optimized builds of this week don't stup down anymore. If I close the
window, I see some msgs on the console until "Clean up ...", then freeze.
Neither window nor console close.

I have "Empty trash at stupdown" (bug 16244) enabled.
annoying..
Severity: normal → major
Keywords: regression
Keywords: nsbeta2
Status: NEW → ASSIGNED
Target Milestone: --- → M17
ben - could you be seeing http://bugzilla.mozilla.org/show_bug.cgi?id=39858?
Lisa, no, it's freezing, not crashing.
Putting on [dogfood+] radar.  We need this to not happen before we can release 
M16.
Keywords: dogfood
Whiteboard: [dogfood+][nsbeta2-]
Testing on today's build I can't reproduce this.  Can you give a more detailed 
description of how to reproduce it?

Ben, it's not freezing. It took time to do empty trash on exit. Especially, if 
you have a lot of mail accounts (server). If you wait for enough time it will be 
fine. I'll try to find a way to hide all windows from the user while doing empty 
trash.
jeff, you're right. I deleted all accounts other than Local Folders (which I
can't delete) and my IMAP server. Now, it "only" needs ~1 min to shut down. 4.x
need ~1s. So, it's not a freeze, but not much better in practice. Adjusting
SUMMARY.
Summary: Freeze at shutdown → Extremely slow shutdown
I have a fix. We stay too long in PR_WAIT(). We should also check for empty 
password to avoid unnecessary server connection. Fix is simple and improve the 
performance a lot.
Whiteboard: [dogfood+][nsbeta2-] → [dogfood+][nsbeta2-] have a fix
Fix checked in.
Status: ASSIGNED → RESOLVED
Closed: 24 years ago
Resolution: --- → FIXED
ben - can you tell me if this is working ok for you now with Jeff's fix? Thanks.

It needs only 5s now. Tnx Jeff.

That's much better, but still not OK. Should I file a new bug?
No, we can reopen this bug if it really sounds terrible. The other thing I can
do is to hide all windows while doing empty trash. This seems overkill to me.
What happen if some other components' shutdown process doesn't wait to have the
window closed?
Ben, how many servers do you have? If the performance is reasonable (2-3 
seconds) for average user I will be satisfied.
REOPENing per jeff's comment.

jeff, this is still with only my IMAP server (machine connected with exclusive
10mbit) (and local folders of course).

IMO, for the final, this needs much more improvement, considering that 4.x needs
1s for 6 accounts. Why are we so much slower than 4.x (just trying to
understand)?

Not sure, how important that is in the short-term. Startup/shutdown performance
is extremely noticable during testing. (I often start Mailnews 10 times is a row
during testing.) Removing dogfood+ for reconsideration.


I don't like the idea of hiding the window. it may hide a possible hang (all
windows closed, but app still active, user starts a new instance, this doesn't
work due to the other running one etc.).
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
Whiteboard: [dogfood+][nsbeta2-] have a fix
My suggestion is early M17.
Since this seems to be better, putting on [dogfood-][nsbeta2-] radar.  Adding 
nsbeta3 keyword to fix in beta3 perf work.  Please let us know if you disagree. 
 And tell us, with current build, how much slower.
Keywords: nsbeta3
Whiteboard: [dogfood-][nsbeta2-]
> And tell us, with current build, how much slower.

Not sure, what you mean with that. Shutdown time now vs. one month ago? For one
IMAP account: Now ~5s and one month ago less than 1s.
Summary: Extremely slow shutdown → Very slow shutdown
I updated yesterday (last update at May 30) and now a real freeze appears: I
click on the window manager close button and half an hour later, Mailnews still
runs. The browser shuts down nicely. Adjusting SUMMARY. Removing PDT evalutions,
so a reevaluation can happen.

Jeff, is that a new bug or is it still the same?
Summary: Very slow shutdown → Freeze at shutdown
Whiteboard: [dogfood-][nsbeta2-]
> Mailnews still runs

rather: ... is still open. No window redrawing, no reaction.
nsEventQueue has been changed lately which may have some effects on this. I 
believe pavlov still has more to land. I'll wait until his checkins before 
taking any actions.
Status: REOPENED → ASSIGNED
Putting on [nsbeta2+] radar for beta2 fix.
Whiteboard: [nsbeta2+][dogfood-]
The freeze is gone now. The previous behaviour shows up again, i.e. ~5s for
shutdown with only 1 IMAP account. Adjusting SUMMARY.

Removing [nsbeta2+]. PDT: You gave that [nsbeta2-] at 2000-05-30 13:17.
Summary: Freeze at shutdown → Very slow shutdown
Whiteboard: [nsbeta2+][dogfood-] → [dogfood-]
[nsbeta2-] assuming freeze is gone now
Whiteboard: [dogfood-] → [dogfood-][nsbeta2-]
adding perf keyword
Keywords: perf
Mail triage marking [nsbeta3-]
Whiteboard: [dogfood-][nsbeta2-] → [dogfood-][nsbeta2-][nsbeta3-]
Summary: Very slow shutdown → Slow shutdown
reassigning jefft's bugs to naving
Assignee: jefft → naving
Status: ASSIGNED → NEW
Clearing very old milestone (M17) in hope of reevaluation.
Target Milestone: M17 → ---
Note: In recent builds (0.8), Mozilla sometimes freezes when closing Mailnews.
Browser dead. Waiting didn't help. New bug?
Keywords: dogfood, nsbeta2, nsbeta3
Whiteboard: [dogfood-][nsbeta2-][nsbeta3-]
QA Contact: lchiang → esther
QA Contact: esther → stephend
off regression radar.  this bug looks pretty old
Keywords: regression
Old Summary: Slow shutdown

New Summary: Slow shutdown, then freeze

Critical severity. Adding keyword hang.
Severity: major → critical
Keywords: hang
Keywords: hang

*** This bug has been marked as a duplicate of 39742 ***
Status: NEW → RESOLVED
Closed: 24 years ago → 21 years ago
Resolution: --- → DUPLICATE
Verified dup
Status: RESOLVED → VERIFIED
Product: MailNews → Core
Product: Core → MailNews Core
You need to log in before you can comment on or make changes to this bug.