337689 - 9.3% Tp regression, 5% Ts on pawn after ThreadManager landing

Status: RESOLVED INCOMPLETE

(Camino Graveyard :: General, defect)

Description

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

pawn jumped from average pageload times of 810ms to 950-960ms following the ThreadManager landing.

Regression range: http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=Camino&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2006-05-10+09%3A00&maxdate=2006-05-10+15%3A00&cvsroot=%2Fcvsroot

Tighter graph: 
http://build-graphs.mozilla.org/graph/graph.cgi?tbox=pawn.office.mozilla.org&testname=pageload&autoscale=1&size=&days=2&units=&ltype=&points=1&showpoint=2006:05:11:20:44:25,959&avg=1

No Mac tinderbox on FirefoxTrunk runs Tp; planetoid (SmTrunk) seems to have only jumped from about 369 ms to 373 ms in the same period (and has been hovering around 370 ms avg for the last 2 weeks), so this seems to be Camino-only.

Ts took a hit, too; we hadn't been as consistent in the days immediately prior, due to the trunk state, I guess, but being generous, it looks like we jumped from about 2500 ms to about 2625 ms avg once things started to stabilize.

http://build-graphs.mozilla.org/graph/query.cgi?tbox=pawn.office.mozilla.org&testname=startup&autoscale=1&size=&units=&ltype=&points=1&showpoint=2006%3A05%3A12%3A02%3A52%3A35%2C2634&avg=1&days=5

It's harder to do a comparison with other trees on Ts since it had just jumped a bunch on planetiod and atlantia prior to the landing (but was back on the way down due to the regression closure work), but Camino/pawn had been very stable around 2500 ms since about 20 April, so the 5% still seems noteworthy.

Comment 1

[Mike Pinkerton (not reading bugmail)]

anyone else involved in that landing that should be cc'd? grr.

Comment 2

[Darin Fisher]

We should probably start by investigating the changes to widget/src/cocoa/.  I haven't had a chance to build camino yet.

Comment 3

[Mark Mentovai]

This might just be an effect of the new Gecko event processing model used for embedding apps.

Comment 4

[Damon Sicore (:damons)]

-'ing.  Would love to see some investigation, but can't block the release as this regression is a bit dated.

Comment 5

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Does anyone believe that this bug is still fixable, or can we close it out as INCOMPLETE?  I'm trying to clean up the regression list for the threadmanager landing a little bit.

Comment 6

[Serge Gautherie (:sgautherie)]

(FWIW, see bug 337516 comment 26)