556410 - Ts regression from Spotlight importer landing

Status: RESOLVED INCOMPLETE

(Camino Graveyard :: OS Integration, defect)

Description

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Bug 549179 comment 16:

> Stuart, any idea why this caused a 3.8% Ts regression on minus, ~1.5% on cb-x4,
> 2.2% on cb-x4-branch, ~0.75% on cb-x1, and ~0.5% on cb-x1 branch?  
> 
> What new code are we running at startup? Or is it just that we now have to make
> UUID-folders and metadata files, whereas before we only made files?

http://wiki.caminobrowser.org/Status_Meetings:2010-03-31:Log#16:13 (can we move creation later?)

Comment 1

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

And/or do it on a delay?

http://mxr.mozilla.org/camino/source/camino/src/bookmarks/BookmarkManager.mm#360 is where we do bookmark site icon fetching on a delay, and http://mxr.mozilla.org/camino/source/camino/src/bookmarks/BookmarkManager.mm#363 is where we write metadata, not on a delay.

Comment 2

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

If we can do it on a delay, that would be nice for 2.1.x, but otherwise, not.

Comment 3

[Stuart Morgan]

We probably could do it on a delay, but looking at the code I'm actually worried now that what we are currently doing is already wrong, and doing a simple just-add-a-delay change could make it much more likely to end badly.

We do the writing on a background thread on startup, while simultaneously continuing to startup on the main thread, including telling everyone the bookmarks are ready, which means edits could start happening--which could trigger writes to the same place.

What we should really do is do both the initial write and the update writes on the same background thread, so they could be queued up and all run in the right order. That's a bigger change though, so I'm guessing it's not going to happen for 2.1.x