mozStorage and Async I/O (was [Bug 408914] Disable sqlite async IO)

Shawn Wilsher

unread,

Mar 9, 2008, 10:30:35 PM3/9/08

to

Per Brendan's urging (and rightly so), moving the discussion that has
been taking place in Bug 408914 Comment 38 [1] and on to the
newsgroups.

Cheers,

Shawn Wilsher

[1] https://bugzilla.mozilla.org/show_bug.cgi?id=408914#c38

Tony Mechelynck

unread,

Mar 9, 2008, 11:08:14 PM3/9/08

to

Please correct me if I misunderstood:

After (diagonally) reading bug 408914, what I understand of the argument
is that async sqlite I/O was disabled for Gecko 1.9 because, at the
moment, no one has the time, the knowledge and the will to do it right,
but that it ought to be re-enabled in some later release because, if
done right, it can result in significant performance gains.

If the above is a reasonably correct summary of the discussion, then
there ought to be a bug somewhere, maybe with a "target" of mozilla2.0
or a "wanted" flag for some not-too-close release, to "enable" async
sqlite I/O again -- and do it right this time. I can't find such a bug,
but I know that I'm notoriously bad at searching the bugzilla database.

Best regards,
Tony.
--
A novice was trying to fix a broken Lisp machine by turning the power
off and on. Knight, seeing what the student was doing spoke sternly:
"You can not fix a machine by just power-cycling it with no
understanding of what is going wrong." Knight turned the machine off
and on. The machine worked.

Shawn Wilsher

unread,

Mar 9, 2008, 11:18:47 PM3/9/08

to

> If the above is a reasonably correct summary of the discussion, then
> there ought to be a bug somewhere, maybe with a "target" of mozilla2.0
> or a "wanted" flag for some not-too-close release, to "enable" async
> sqlite I/O again -- and do it right this time. I can't find such a bug,
> but I know that I'm notoriously bad at searching the bugzilla database.

No bug on file as of yet.

Cheers,

Shawn Wilsher

smaug

unread,

Apr 7, 2008, 7:17:31 AM4/7/08

to comra...@gmail.com

So has anyone tested this properly with non-local profile directories?
https://bugzilla.mozilla.org/show_bug.cgi?id=408914#c3

https://bugzilla.mozilla.org/show_bug.cgi?id=408914#c26
as Brett, I'm also extremely worried about the change.
Many companies and universities etc have user directories
on (possibly quite slow) network drives.
We don't want those to ditch FF3 because of possible UE/UI performance
problems.
So has anyone tested this properly?

smaug

unread,

Apr 7, 2008, 12:24:07 PM4/7/08

to

The first comment in the bug is
"Disable our async IO; there's a theory that it was causing corruption
in some cases, so let's see what the perf hit is if we just turn it
off."

Was that theory proved to be true?

Mats Palmgren

unread,

Apr 7, 2008, 8:19:04 PM4/7/08

to

smaug wrote:
> Many companies and universities etc have user directories
> on (possibly quite slow) network drives.

There are several reports of temporary freezes using *local*
profiles too, eg:
https://bugzilla.mozilla.org/show_bug.cgi?id=421482#c7

/Mats

Shawn Wilsher

unread,

Apr 9, 2008, 11:56:46 PM4/9/08

to

On Apr 7, 12:24 pm, smaug <sm...@welho.com> wrote:
> Was that theory proved to be true?

The number of corrupted cookie/downloads/places bugs that I've seen
has gone to zero since that landed. So not proven, but evidence
suggests that.

Shawn Wilsher

unread,

May 4, 2008, 12:22:26 AM5/4/08

to

On Mar 9, 10:30 pm, Shawn Wilsher <comrade...@gmail.com> wrote:
Note: Bug 429986 [1] should be of interest to folks.

Cheers,

Shawn

[1] https://bugzilla.mozilla.org/show_bug.cgi?id=429986