528505 - hang after starting 3.0rc1

Status: RESOLVED WORKSFORME

(MailNews Core :: Backend, defect)

Description

[Wayne Mery (:wsmwk)]

Attachment: windb hang stack #1

solid hang after starting 3.0rc1 on *first* start with existing profile - profile previously used with 3.0 builds.  last build used with profile was 20091112.  So it seems odd that I should hang using RC1.  master password prompt is up, but unresponsive.

Tried to reproduce by setting mail.startup.enabledMailCheckOnce to false, on the assumption that it's related to "What's New" page, but the page doesn't come up, and no hang.  So another possibility is it may be extension related (but not theme, I use the default).

second and subsequent startups don't solid hang, but aren't perfect. ref: bug 528499.

hang stacks attached, both from same startup of thunderbird

Comment 1

[Wayne Mery (:wsmwk)]

Attachment: windbg stack #2

Comment 2

[Wayne Mery (:wsmwk)]

I doubt I'll have time to work more with this today, unless there's something really simple for me to try.

Comment 3

[David :Bienvenu]

Looks like sqlite is blocking (perhaps trying to do the soft-update of the schema?). Or perhaps finishing committing transactions after a previous crash or something. Asuth would know much better.

Comment 4

[Wayne Mery (:wsmwk)]

Since my prior run was with 20091112 build, should it need to do any updating at all?

Comment 5

[rsx11m]

I have just migrated a profile from Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.6pre) Gecko/20091110 Shredder/3.0pre to Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9.1.5) Gecko/20091112 Thunderbird/3.0 without seeing any problem. I let it finish indexing with the old version, sent myself a message to make sure that something to index is there, then restarted with 3.0 RC1 with the What's New page and a new message indicated, no delay. Following 2nd start-up without no new mail was ok as well.

Comment 6

[rsx11m]

(In reply to comment #4)
> Since my prior run was with 20091112 build, should it need to do any updating
> at all?

When I went back to the nightly build (i.e., to 1.9.1.6pre from 1.9.1.5),
Gloda started reindexing from scratch without apparent reason. Maybe a change
of the milestone is contributing in whatever way?

Comment 7

[Mark Banner (:standard8)]

(In reply to comment #6)
> (In reply to comment #4)
> > Since my prior run was with 20091112 build, should it need to do any updating
> > at all?
> 
> When I went back to the nightly build (i.e., to 1.9.1.6pre from 1.9.1.5),

That's the wrong version numbers to include in this case, you want the dates, so "to 20091110 from 20091112".

> Gloda started reindexing from scratch without apparent reason. Maybe a change
> of the milestone is contributing in whatever way?

In between those dates asuth checked in a change that soft-updated the version of the schema. So going back again will kill the database and restart indexing.

Comment 8

[rsx11m]

Thanks, that explains my observation (though not Wayne's case).

Comment 9

[Wayne Mery (:wsmwk)]

my problem profile is backed up with mozbackup if we need it.

Comment 10

[Wayne Mery (:wsmwk)]

(In reply to comment #3)
> Looks like sqlite is blocking (perhaps trying to do the soft-update of the
> schema?). Or perhaps finishing committing transactions after a previous crash
> or something. Asuth would know much better.

Andrew, is this something worth attempting to reproduce at some future date? Or is it more than likely something already addressed somewhere?  (And I can toss the profile)

Comment 11

[Andrew Sutherland [:asuth] (he/him)]

Both of your hang stacks are indicative of builds with an unpatched mozStorage.  The main thread joined on the async thread which was doing something long-running.  Toss the profile.