MailNews Performance test results for Win32, Red Hat Linux 6.1 and MAC
OS have been updated and can be viewed at
http://www.mozilla.org/mailnews/ under Quality Assurance.
Tested on comm build 2000-08-16-08, M18.
Thanks to:
-
srilatha@netscape.com (sustaining eng.) for diving into some of Alecf's
easy nsbeta3+ bugs
-
Thanks to dougT for his support on bug 46320 and to mscott for his quick
code reviews for activation related bugs.
Issues
-
Very high nsbeta3 incoming rate
-
Have a hack fix for 31263 remembering splitter position in mail windows
- have to talk to waterson about persistence.
-
Have had trouble getting the SecurID contacts in place. Hopefully
today's conference call will resolve that.
Vacations & Sabbaticals!
-
JeffT 7/10-9/5
-
SSpitzer 7/5-8/26
-
RHP 8/18-8/27
-
Alecf in Australia 8/21-9/8
MailNews
| Date |
Open nsBeta3+ P1/2 |
Open nsBeta3+ P3/4/5 |
UnTriaged nsBeta3 Nominees |
Fixed nsBeta3+ |
Fixed non-nsBeta3+ |
| 7/31 |
0 |
0 |
630 |
10 |
46 (and 8 b2+) |
| 8/7 |
0 |
135 |
57 |
33 |
47 (and 9 b2+) |
| 8/14 |
0 |
146 |
43 |
56 |
16 |
| 8/21 |
69 |
62 |
70 |
43 |
6 |
Profile Manager
-
Bug 46320 (moving profile registry to HOME) was landed last week. That
is going make people's life easy by moving registry into the location which
is not usually restricted by any previleges.
CCK
-
small amount of bug work going on
Highlights
- Finished first round of API reviews.
Issues
-
Architecture/Embedding bug triage status:
-
up to 13
untriaged beta3 bugs!
-
assigned to attinasi, av, beard,
danm, leaf, morse, mscott, pnunn, rusty.lynch, and thayes
-
It has been 4 weeks now... please
take a look at these.
-
28
of the bugs we consider essential for embedding
have been nsbeta3-'ed (27 last week, so maybe we've stabalized this)
-
102
bugs have been accepted for beta3 (nsbeta3+'ed) and are still open
- Footprint meetings have been going well, and we're knocking off bugs
we've placed on the list. However, we still don't have a sense that we're
making a significant dent in the overall footprint of the app. We need to
get more people looking at bloat to detect other opportunities for optimization.
Engineering Status
-
waterson
-
Checked in fixes for
-
48086, regression from content model teardown optimizations
-
44480, document.width needs to measure <body> element
-
47451, \n and \r not treated as whitespace
-
49122, fix image map <area> frame manager monkey business
-
29641, turn off "backoff" count (which was strangling layout)
-
46846, doc charset issues with image maps and <link> elements
-
Awaiting for review on
-
47843, "style:overflow" on <body> crashes
-
Debugged cache corruption problem (49108) with gordon, worked up some defensive
code to at least avoid crashing when we get a corrupted cache. Patches
in the bug, currently assigned to neeti.
-
Triaged the dozen or so new bugs I've received; up to 10 [nsbeta3+] bugs,
all prioritized according to the *New* New PDT Guidelines.
-
Helped Edward Kandrot get started on leak fixing (49428).
-
No word from IBM re: bidi text. I suspect that they are pretty near to
being ready to go with their changes on their branch.
-
vidur
-
rayw
-
waldemar
-
Worked on JS2 language specification.
-
Prepared for meeting with Microsoft next week.
-
rpotts
-
just got back from vacation -- no status
-
jband
-
Spend Friday at the mozilla developer conference.
-
Worked on the same collection of bugs as the previous week. No great closure.
I plan to pull a lot of that together in the coming week and get a bunch
of stuff checked in.
-
warren
- Finished code for rearchitecting streams. Awaiting review.
- Fixed more problems with hash table string keys.
- Worked on jar packaging.
- Added name attribute to nsIRequest for embedding APIs.
- Added error checking to chrome registry.
- Worked on extending properties file format to support multi-valued
properties (needed for Ray's component registry work, and Hyatt's xap file
idea), but got push-back from i18n guys on changing the format.
-
dprice
-
scc
-
string work, and fallout from the new multi-fragment stuff
- got substrings working again, which meant refactoring some iterator and fragment code
-
supported vidur and jst's effort to convert the DOM APIs
-
wrote a nsPrintfCString, which replaces the functionality of AppendInt et al,
but, as a first class string object, provides even more flexibility
-
added some conversion routines that convert the encoding of a readable into a writable destination;
can't have a multi-fragment destination yet, but this'll do for now
-
helped with the new nsPromiseFlatString class
-
helped jst with compiler issues and debugging
-
helped resolve the issues with nsStringKey
-
reviewed, tested, and/or checked-in other peoples patches
- bug #47244: the new tag doesn't work, so we're not ready for jcoopers's patch
-
wrote documentation
-
posted answers to `which string classes should I be using';
I'll get this local and have it be part of the documentation I'm slowly building
-
more string `how to' answers which will become part of the
complete documentation
-
attended meetings
mozilla.org staff meeting (1 hour)
-
kandrot
-
Last weeks' status...
- Learned more of the tools.
- Talked with people about where performance issues might lie.
- Worked with Wade to try to get Linux refCounter to work.
- Learned COM.
- Learned basics of XPCOM.
- Set up Win2k machine from home to use as build machine until AOL machines arrive.
- Meet with people to get a picture on how source code works.
- Learn about bugzilla and entered a bug.
Summary
-
The XPToolkit team resolved 79 bugs in the last week, fixing 57 of these, including 48 '+'. For details,
see our
resolved
bug list.
- Prioritized our triaged bugs to separate must-fix from may-fix; we currently have only 26 bugs that we have to fix for nsbeta3. See Beta Stoppers section for details.
Highlights
-
Mike Pinkerton (pinkertoni)
- fixed mail d&d by turning off mouse capture when a drag starts (48250)
- retooled tree d&d feedback to be less noisy when dragging over containers, use the style system more for feedback, and support trees where drags are not allowed between rows (folder pane in mail) (43421
and 47105)
- fixed bugs where multiple submenus could appear at the same time, and the side effect that when that happened clicking menus would do nothing. thanks to dean_tessman@hotmail.com for a starting point to fix both bugs. (bug #s 48989 and 28309)
- paste into text areas on win98 only no longer shows funny characters (48291)
- made first time form submission faster, but still not fast enough. we know what the problems are now, at least (bug 44514)
- profiled window refresh, discovered it's 50% painting, 50% focus, and we're doing a lot of weird things (many of them XP)
-
Chris Saari (saari)
- Partial fix for 46938 (plugin is not getting key events), on Windows only.
-
Daniel Matejka (danm)
-
Eric Vaughan (evaughan)
-
Stuart Parmenter (pavlov)
- fixed 7 nsbeta3+ bugs (resolved 8 total nsbeta3+ bugs)
- 26502 - resolved linux performance bug.. performance is pretty good on linux now and any new issues should be filed as separate bugs.
- 16710 - remote mozilla x11 display is very slow. remote mozilla performance is pretty much acceptable now
- 49530 - copied modern skin to the new 'blue' skin so that while the modern skin is being redone (again?), we have a stable skin to use.
- 46729 - fixed pasting of things like smartquotes as plain text
- 49537 - fixed leaking of XGC in the GC cache
- 49555 - fixed leaking of a GC in nsImageGTK::Draw()
- 48839 - CSS3 'spinning' cursor uses wrong icon
-
David Hyatt (hyatt)
- Fixed 23 bugs, including 18 nsbeta3, several crashers.
-
Dan Rosen (dr)
- Coming up to speed on XPCOM
-
Brian Ryner (bryner)
- Fixed crash if you try to mousewheel scroll an empty tree (48306)
- Filepicker fixes:
- Checked in contributor patch for 49089 (Enter key broken)
- 48311 (Make * filter really show all files)
- 47984 (crash when starting in an invalid directory)
- Fixed window position persistence on Linux (42345)
Lowlights
Priorities
-
pinkerton
-
work with simon, beard, and saari to speed up mac stuff (our memory allocator could use tuning)
-
saari
-
hyatt
- 44093 scrolling folder pane [@ nsXULTreeOuterGroupFrame::Fin
- 48036 Columns in thread pane are improperly aligned with their con
-
evaughan
- 35681 HTML is sized too wide
-
pavlov
- 46409 - crash/hang mozilla when you copy in nedit
- 45074 - crash when you specify bad command line parameters
-
danm
- 2020 Crash exiting browser
- investigate 49358, which is shaping up to be a subtle problem
underlying several bugs .
- 49007, windowshaded window problems on the Mac
-
dr
- 47395 (p3, beta3+), make nsXULKeyListenerImpl::HandleEventUsingKeyset not suck
-
bryner
- 42109 crash at shift-clicking a title in the messages list
-
47795 Linux filepicker is slow to load large directories
Dogfood
-
Fixed two dogfood+ bugs in the last week:
- 49066 Crash starting up mail when we try to bring up the password
- 48126 60k of leaks added to smoketest
-
No outstanding dogfood+ bugs.
Beta Stoppers
We have now prioritized our triaged bugs to indicate which are must-fix,
and which are only may-fix. We are committing to fix for nsbeta3 only
those bugs that we have assigned, accepted, plussed, and marked with priority
P1-P3. We currently have 26 bugs in this category, 1 in bugscape and 25 in bugzilla. We also have 3 untriaged crashers which will almost certainly require fixing.
The 18 bugs that we plussed with P4-P5 priority
may still be fixed (in priority order), but we're not committing to fix
them. We'll be eliminating these entirely over the next 2 weeks as the risk
starts to outweigh the benefit.
Bugs on our list that have not been triaged by us, and accepted (i.e.,
they are status 'new'), may have been marked nsbeta3+ by others, but we
have not committed to fixing them (although we probably will commit to most
when we do triage them). We don't have any of these right now, but have
dealt with a signficant number over the past few weeks.
Our glide path still seems reasonable, and we hope to start bouncing off
zarro individually soon, and be able to help other teams out with other
difficult bugs and performance problems. In most cases, this will take priority
over any P4-P5 bugs we may still have.
Risk
The main risk we face now is getting more new showstoppers than we can fix.
If you have any bugs that you think may really be toolkit problems, please
let us know as soon as you can. If you see new problems, please report
them right away.
Issues
Escalation
People
- Eric Vaughan should be back from vacation on Tuesday.
Summary
-
High nsbeta3 count and just getting higher!
-
neeti back from vacation!
Dogfood and Risks
Beta Stoppers
Highlights
-
neeti
-
pnunn
-
Bugs Closed: bug#48910, bug#48665, bug#37241, bug#20939, bug#12977,
bug#46236
-
Bugs Reassigned: bug#48790, bug#48114
-
I concentrated on image cache/necko cache issues this week:
-
Attended the perf/foot meeting and worked with Syd and Gordon to pin down
where problems are occurring.
-
Verified cache symptom Gordon described in current tree on NT.
-
Went to moz developer's meeting for a short while.
-
Bug Work:
-
Started to piece together what I had for bug#13718. Needed by editor
folk.
-
Looking at effects of image requests from images in local files. (nsFileChannel)
w/r to 304's.
-
gordon
-
Triaged bugs for nsbeta3
-
Investigating cache problems with Imagelib. ImageCache and Necko Cache
interactions.
-
gagan
-
Bugs resolved: Bug 48200.
-
Triaging bugs for the whole team.
-
rjc
-
Fixed bug 47927
-
Attended meeting in Boston.
-
ruslan
-
Fixed the hotmail.com bug.
Lowlights
-
ruslan
-
no status from ruslan this week as well.
Issues/Escalations
-
gagan
-
I am just getting swamped with process and not getting enough cycles to
work on the bugs. Either we need to find my evil twin brother or simplify
our process that emphasizes more on getting work done instead of planning,
nominating, triaging, etc.
Summary
-
We are down from 77 to to 47 nsbeta3+
bugs as a result of triaging the list and checking in fixes in hand.
We need to fix 3 bugs per week per engineer for the next four weeks if
we want to get to zarro boogs by 9/14. We'll monitor the bug count
twice a week and try to stay on a reasonable glide path.
People
-
Peter Van der Beken from Belgium is on the Netscape campus. He is
hacking on XSL. Come down to the Gecko area to say hi!
-
Tom is on sabbatical till August 25th.
Accomplishments
- Harish Dhurvasula (harishd)
-
Triaged my bug list & Clayton's list ( 6 bugs )
-
Checked in fixes for 4 nsbeta3+ bugs ( 8 bugs )
-
Token based arena cache is reporting huge bloat numbers!!! ( should be
the other way around). Working on it.
- Heikki Toivonen (heikki)
-
Resolved 4 bugs: 49005 (WONTFIX), 45368 (WORKSFORME), 36062 (FIXED, was
one on joki's patch), 48770 (WORKSFORME)
-
nsbeta3+ bugs: 46085 (waiting review), 48277 (still can't reproduce, considering
WFM), 30878 (partial fix, waiting comments, thinking if this should be
futured after all), 8927 (partial fix, working on it), 48488 (after more
studying seems like there is nothing to fix).
- Johnny Stenback (jst)
-
Worked on the DOM API string changes, almost ready to land
-
Fixed nsbeta3+ bug 43327. Made window.history act like an array to be compatible
with 4.x, also added proper security checks for DOM array access.
-
Fixed nsbeta3+ bug 44002. The .complete property on image elements in mozilla
is a boolean property but mozilla was incorrectly returning the value 2
in stead of one. This caused scripts that rely on img.complete == true
when an image has loaded to break (since 2 != true).
-
Fixed nsbeta3+ bug 30534. Correctly report what DOM features mozilla supports
if asked through DOMImplementation::hasFeature() or Node::supports().
-
Fixed nsbeta3+ bug 45680. Cloning an HTML element with a style attribute
didn't clone the style declaration in the element so changing the style
on the clone resulted in a style change for both elements.
-
Fixed nsbeta3+ bug 46966. Changed the exception thrown when a DOM Attribute
node is modified (children added or removed).
-
Fixed nsbeta3+ bug 47420. Don't throw a DOM exception when accessing a
nodemap by index even if the index is out of range, simply return null
instead.
-
Fixed nsbeta3+ bug 46190. Clicking on links that point to anchors with
whitespace in it ("http://whatever/index.html#some name") wasn't working...
-
Fixed nsbeta3+ bug 31539. Adding support for the 4.x DOM property 'text'
on links.
-
Fixed nsbeta3+ bug 42670. Stop the JS (DOM) error reporter from recursing
to death if there's a JS error in an onerror handler.
-
Fixed a few memory leaks in nsHTMLTableElement.
- Nisheeth Ranjan (nisheeth)
-
Reviewed Tom Lendacky's (toml@us.ibm.com) P3P related changes to the document
and the pres shell
-
Gave a talk on "Open Source Development, Mozilla.org, and the Gecko Layout
Engine" to the summer interns
-
Triaged XML/DOM team bugs. Beta3+ bugs are down from 77 to 47.
Beta 3 nominees are down from 40 to 7.
-
Met with Keith Visco, Peter Van der Beken, and Vidur to discuss status
and future plans for XSL.
-
Talked with Radha and Rick Potts about storing document.write generated
documents in session history and using form post data as a key into the
session history entry. Set up a meeting for Monday with Brendan,
Vidur, Gagan, and Rick to explore this further.
Plans for next week
- Harish Dhurvasula (harishd)
-
Will be attending the meeting on Strict DTD
-
To have my token cache changes reviewed and checked in after resolving
the bloat numbers
-
Work on NOSCRIPT ( bug 7670 ) content
-
Triage my bug list.
- Heikki Toivonen (heikki)
-
Continue with nsbeta3+ bug fixes and bug triage.
- Johnny Stenback (jst)
-
I'm hoping I can finally land the DOM API string changes.
-
Continue working on my remaining nsbeta3+ bugs.
- Nisheeth Ranjan (nisheeth)
-
Get focal reviews signed.
-
Ongoing triage of beta3 nominated and untargeted bugs
-
Integrate latest expat release into Mozilla and start passing up information
about ID attributes to the DOM
Lowlights
Escalations
We have obtained an IBM ftp location to place our nightly builds, and are
working towards getting them available early next week.
We're not getting a lot of outside help reporting bugs, so our main focus
has been finding problems with the browser so we know what to fix.
We're concentrating on overall stability.
The Bidi code is on the branch IBMBIDI_20000809_BRANCH.
We are currently working on a crash that only happens on the release build.
Once we get that fixed, we hope to be able to land the stuff in #ifdefs.
Previous Updates
|
