521008 - 11% Ts regression from landing of Windows 7 Jumplists

Status: RESOLVED FIXED

(Firefox :: General, defect)

Description

[Marco Bonardo [:mak]]

I'm splitting this out from Bug 520837 since the 2 codes are separate and caused regressions at different times.

looks like Jumplists don't check for wintaskbar availability before trying to Cu.import the module, Aeropeek does that. So we import the module just to discover we are not on Seven and bail out.

There could be other reasons for the regression though.

Asking blocking p1 and assigning to Jimm per mirroring Bug 520837

Comment 1

[Marco Bonardo [:mak]]

11.43% on XP, 8.12% on Vista

Comment 2

[Marco Bonardo [:mak]]

hm, jumplists did not land on 3.6 from what i can see. clearing blocking req.

Comment 3

[Jim Mathies [:jimm]]

I'd rather not have a Ts regression on 7 either, so maybe there is a way to delay the load of all this since it's not central to startup of the browser.

Comment 4

[Marco Bonardo [:mak]]

well sure, but at the same time why losing time on other OSes, even if we remove this from Ts it is still useless work. Aeropeek and future downloads progress are doing that, so i think jumplists can do the same (and maybe use the same WIN7_FEATURES define that is added there for that purpose)

Comment 5

[Jim Mathies [:jimm]]

Just pushed a fix for a bug where the jsm wasn't being imported due to a bad filename. It'll be interesting to see how that effects this.

http://hg.mozilla.org/mozilla-central/rev/bea32394861e

Comment 6

[Jim Mathies [:jimm]]

Attachment: win7 trap for import and delayed update v.1

Comment 7

[Jim Mathies [:jimm]]

Attachment: sysinfo trap for import and delayed update v.1

Comment 8

[Jim Mathies [:jimm]]

I've pushed both of these to try to see what effects they have. Using available might be slower since you have to initialize the taskbar object to get the result.

Comment 9

[Jim Mathies [:jimm]]

Went away after the bad import path patch (bug 518666) landed.

Comment 10

[Jim Mathies [:jimm]]

(In reply to comment #9)
> Went away after the bad import path patch (bug 518666) landed.

See bug 521104 for details on various issues related to bad import paths.

Comment 11

[Marco Bonardo [:mak]]

out of curiosity, did one of the two patches above show any improvement on tryservers? i'd expect them to bring an improvement on XP, since there the module is completely useless but it is still imported.

Comment 12

[Marco Bonardo [:mak]]

nevermind i've seen the patch in bug 518666. even if all of these patches are related it's hard to notice patches on an already fixed bug :)