May 23 Gecko 1.9 Planning/Status Meeting
Damon Sicore
PDT. Here's a link to the meeting agenda:
http://wiki.mozilla.org/Firefox3/StatusMeetings/
2007-05-22#Gecko_1.9_Meeting
Before the meeting, please add any agenda items as well as any Gecko
1.9-related progress to this page/section.
Meeting Details
* Wednesday, May 23, 11:00 am PDT
* 650-903-0800 x91 Conf# 217 (US/INTL)
* 1-800-707-2533 (pin 369) Conf# 217 (US)
* join irc.mozilla.org #granparadiso for back channel
Cheers,
Damon
fantasai
Damon: Welcome to Gecko 1.9 meeting
<gnubeard> http://wiki.mozilla.org/Firefox3/StatusMeetings/2007-05-22#Gecko_1.9_Meeting
Trunk Instability
-----------------
Shrep: Do we know what sort of things are causing the recent trunk
instability?
?: no
dbaron: I don't think I've actually quit the browser such that I
could get leak numbers
dbaron: I'm crashing probably once every day or two
dbaron: a different thing every time
Talk about some particularly unstable builds.
Perf Regressions
----------------
dbaron: The performance regessions on the windows boxes were tweaks
to the tinderboxen, not actual regressions.
sayrer: There was a real regression on windows. There was a 10%
regression on the tp.
sayrer: We had a bit of a win with the string changes, but that was
lost. This is separate from the tinderbox issue.
dbaron: So the only possibility I see is that the task manager code
caused a windows-only regression.
dbaron: Somebody should take builds before/after and make a test.
<gnubeard> See 381537
https://bugzilla.mozilla.org/show_bug.cgi?id=381537
Shrep: So we need to do two things ... checkin window ...
Shrep: If we have a perf regression and it's not clear whose fault
it is, then we need to ...
dbaron: Someone needs to run the tests before and run the tests
after to make sure there's actually regression
sayrer?: I'm not sure that there's a tp hit, I'm not sure that there's
not. So I agree with David that we should run baseline.
sayrer?: I'll run the tests this time, but I don't think in general
that objecting to perf regressions should create work for
that person.
People analyze the graphs, some on the call and some on IRC
<gavin|>
http://build-graphs.mozilla.org/graph/query.cgi?testname=pageload&units=ms&tbox=bl-bldxp01_HEAD&autoscale=1&days=7&avg=1&showpoint=2007:05:23:10:31:46,591
is the relevant graph
<gavin|> the average is down from what it was before
<johnath> yeah - that graph shows a ~1% spike post, but it's hard,
because the numbers seem to have about +/-2% noise from
run to run
<Pike> wasn't the reboot *after* the regression? I recall that
aravind did that to test if anything on the box was causing it
<Pike> repeating myself for dbaron, wasn't the reboot *after* the
regression? I recall that aravind did that to test if
anything on the box was causing it
<gavin|> not quite down to what it was immediately after the first
drop, but I think that's within te noise threshold
<johnath> gavin|: right
<sayrer> gavin, I don't 580 to 600 is within it
<dbaron> Pike, there was a bit of orange before the regression
<dbaron> Pike, or was it red?
<Pike> might have been a red on the branch, not really sure
<dolske> Is MV fading in and out for people on the call?
<sayrer> yes
<gavin|> sayrer: previously the numbers varied between ~580 and ~615
<johnath> sayrer: the avg went from 580 to 585(ish)
<Pike> dolske: it's horrible for me
Blockers
--------
shrep: Based on our current fix rate we will never ship.
shrep: The best thing we can do is triage what are really blockers
shrep: Anyone who can help with these blockers, please help out.
shrep: I'm getting really scared about the number of blockers .. ?
...
shrep: Jonas?
<roc> the phone connection is really bad
<dietrich> yep, just got worse
something about dogfood and shipping now vs later
Updates
-------
Damon: vlad, are you on the line? or roc?
roc: I'm here
Damon: Do you want to talk about..?
roc: It looks pretty good
roc: should work most of the time
roc: ... nightly builds ... textframes
roc: There's a bug in the nightly builds ....
Damon: Layout update?
Damon: Content update?
Jonas: working on regressions
Damon: OSX?
Josh: So we made some progress on the blockers list, and we're ready
to maybe turn on form controls.
Josh: I tried it once, but it caused some reftest failures, so I'm
working through those.
Josh: Hopefully we'll make it in time for a5, if not then a6
Damon: That's a risky change.
Damon: Yesterday we talked about this for Firefox. We concluded that
it probably wouldn't be risky crash-wise.
?: It has more potential for causing layout problems.
??: The problems we're seeing are mostly about sizing.
<dolske> grumpy dolske: with all the recent trunk problems, and it
being this much closer to the A5 freeze, do we really want
to rush this in?
Josh: There's one bug where the size of a particular widget is too
wide. Hopefully I can fix it for a5, but I don't think it's
bad enough that we want to hold off from a5.
Jonas: Form controls are heavily used on the web, so it's a lot of
pages that might be affected.
dbaron: It looks like that bug we were talking about last week.
Josh: I'm pretty sure it's a different bug.
Josh: It's definitely not the ? width bug.
Damon: Don't want to discount awesome work done on the mac, if
you've thought about the risk and made a decision.
Josh: If people think it's too risky, I won't take it personally.
Josh: It's a simple switch. Easy enough to turn it off.
Damon: Mac OS X 10.3 report
Josh: I suggested on a newsgroup that we should drop for 10.3 panther.
Josh: Looking at our schedule, I don't think we have the resources to
do the things we want to do for tiger and leopard without
dropping support for 10.3.
Josh: I think we should focus on where the vast majority of users
are: 10.4 and 10.5
?: This is only affecting the 1.9 release.
??: I think it's a reasonable policy to drop support for older OSes
if we have something else, like FF2, which will be supported for
a reasonable amount of time
Josh: We're supporting 10.3 for 6 months longer than Apple is
supporting 10.3
Josh: I don't think I made the point strongly enough. Of course it
would be nice if we supported 10.3, but we can't do everything.
no objections
Damon: Build Tools update
Ted: Benjamin couldnt' make it
Ted: He made a python script to pull NSS/NSPR/etc and check it in,
so you can build the mozilla2 repo
* fantasai didn't catch all that; too fast
<luser> it's pretty much all on the wiki :)
Damon: Any other issues?
Meeting closed.