Mozilla

Since we no longer have Friends of the Tree I'd like to do a quick Mozilla Community QA and Testing status report to thank some folks for an amazing week of Unconfirmed bug triage. The last few weeks have seen the total Unconfirmed bug count drop from over 600 to under 100. A special thanks to ksosez@softhome.net for taking a real leadership position in incoming bug triage. The numbers below are for triage of Unconfirmed bugs (changed from Unconfirmed to Resolved or Confirmed) in the last 7 days.

• ksosez@softhome.net (Keyser Sosez): 105 [You Totally Rock!!]
• dark@c21.net (R.K.Aa): 40
• bzbarsk@mit.edu (Boris Zbarsky): 25
• hidday@geocities.com (Fabian Guisset): 20
• st.n@gmx.net (Stephan Niemz): 17
• matti99@gmx.de (Matthias Versen): 16
• walk84@usa.net (Stephen Walker): 11
• davidr8@home.com (Jesse Ruderman): 9
• bhart00@yahoo.com (Bradley Hart): 9
• hwaara@chello.se (Håkan B. Waara): 8

I'm sure I missed a few people here (and I was only looking at unconfirmed triage). If you've been helping to maintain sanity in the database and you've got some stats you'd like to see highlighted please email me and I'll add it to future reports.

We are tracking memory usage on daily smoke testing results ftp://ftp.mozilla.org/pub/leak-reports/ daily_memtests/memtest_results_daily.html and even better than that, avg. memory usage per page load is on the way down! ;-) ftp://ftp.mozilla.org/pub/leak-reports/ daily_memtests/slope.png

We should also see a good jump lower on today's smoke test results after the landing of radha's session history improvements on Friday.

Summary

• gagan
  • great progress on cache design
  • Pavlov joins our group as lead on imagelib

Highlights

• darin
  • Ready to land AsyncWrite changes.
  • Triaged all HTTP bugs; developed plan of attack with neeti.
  • Cache design refinements.
• neeti
  • bugs investigated, reassigned 64621, 64857, 39046, 46993, 20195 65569.
  • Investigated bugs 53100, 56633, 56523.
  • Currently working on redirect bugs 61693
• gordon
  • Held design review for cache design changes.
  • (valeski, mscott, dveditz, dougt, saari, pavlov, darin, chofmann attending).
  • Created NECKO_CACHE_20010119_BRANCH.
  • Checked in cache interfaces.
  • Invalidate Bugzilla bug #65639
• dougt
  • Investigate bug 65177. Found that it was caused by a known problem in the htmlparser.
  • Implemented FTP EPLF in Mozilla. Patch is in 49958.
  • Added pref code so that PSM could disable UI if needed for embedding (65623)
  • Hacked fix for dots in a relative path. 53152
  • Leading the effort to make nsIChannel bidirectional.
• pavlov
  • Joined Gagan's team to start work on imagelib.
  • Produced timeline and task list covering the changes required to imagelib.
• gagan
  • pavlov joins our group! we finally have a lead on imagelib and he has already started coding!
  • picking up the threads for necko/imglib. Thank to gordon for the excellent coverage while I was gone.
  • task planning in details. Getting finer numbers and tasks for our plans for the next version and beyond.
  • helping dougt with his nsIChannel bidi changes.

Lowlights

• dougt
  • Reports of FTP hanging. 65220
  • Found terrible bug in nsIChannel which prevents overlapped IO.
  • Some change happened in xpcom component loading which broke stream converters. I asked all the suspends, and they denied involvement. I need to spend time now fixing this potentially involved bug.
• gagan
  • lots to do for cache and imagelib and we are running out of precious time real fast.
  • learning the same old painful way of using MSProject. How I wish there was an alternative...

Issues

Plans for the next week

• darin
  • Land fixes for key HTTP feature bugs (eg. redirect looping cleanup).
  • Land AsyncWrite changes.
  • Work with dougt on nsIChannel cleanup.
• neeti
  • Work on redirect bugs
  • More bug triage
• gordon
  • Hold high level design review. (hoping to get rpotts, vidur, brendan, waterson, others).
  • Work on Implementation of nsCacheManager.
• dougt
  • nsIChannel and related interface changes.
• pavlov
  • Finish up technical documentation for imagelib plans to go along with the task list so that people can begin writing code
• gagan
  • finish up on dougt's changes and help land that
  • planning for imagelib in details
  • help gordon with cache development

Summary

• Tooltip embedding APIs are now working.
• The XPToolkit team resolved 18 bugs in the last week, fixing 7 of these. For details, see our resolved bug list.
• Coming up to speed on MSAA & refining plan/design. Trying to write a simple AA server. Looking for example source code for one, or someone who can help us learn how to write one faster.
• Window API tasks are still blowing out our mozilla0.8 milestone budget, need to split them up again, reassign.

Highlights

• Mike Pinkerton (pinkerton)
  • fixed a clipboard leak pav showed me
  • finished tooltip embedding api's and they work in the mac embed sample with balloon help
• Chris Saari (saari)
  • got branch and trunk plugin changes synched up
  • 65129 keyboard events go to window while playing game on a page
  • 58957 Mac plugins do not receive key events
  • 54406 Can't tab past a link which contains a hard or soft line break
  • lots of ImageLib chatting with pav, gordon, darin
• Daniel Matejka (danm)
  • Heh. Well, I have that window manager component I mentioned last week as one of my embedding tasks building on Windows. It's not yet completely hooked up in the application, either. I do have an API that's pretty close to final.
  • Before Monday, I'll have completed and disseminated a disturbingly large task: describing and fleshing out what's on my plate for embedding. Scheduling follies take two to follow.
  • 65262 fix bookmarks and history window size/pos/state persistence
• Eric Vaughan (evaughan)
  • installed MSAA sdk and analyzed how IE does accessibility.
  • hyatt and I created a basic design for Gecko.
• David Hyatt (hyatt)
  • Investigated IE's MSAA support & started outlining plan/design.
• Dan Rosen (dr)
  • Participated in review of XSLT for inclusion in moz0.9/1.0
  • 55115 convert chrome registry to use nsIFile (not nsFileSpec)
  • 50121 --help doesn't show --gtk options (maybe dependent bugs 36532 and 54947 also)
  • May have successfully fought for standard APIs for XPath 2.0 (so that XPath can be used portably in scripts).
• Brian Ryner (bryner)
  • got much psm 2.0 work checked in

Lowlights

Priorities

• pinkerton:
  • get all this code sr'd and landed. I've got a lot
  • find out how to do mac image --> pict conversion for mac image copy
  • Help DanM with Window API tasks.
• saari:
  • Unblock rods and dcone printing tasks with event loving
  • mozilla 0.8 bugs
  • imglib2 work (GIF, JPEG decoders off the floor)
• hyatt:
  • Design/implement MSAA support for Gecko.
• evaughan:
  • Get a simple MSAA server running inside of win-embed.
• danm:
  • Describe and divvy up my remaining embedding tasks.
  • Finish window manager component.
• dr:
  • Help un-doom danm (embedded window APIs and impl)
  • Find a better way to write XML besides emacs, and get back to work on XBL and XUL specs the following week.
• bryner:
  • Help DanM with embedding window implementation.
  • tree performance, especially in ftp

Risks

• Writing an MSAA server is still a largely unknown task, need to solidify plan/design. Starting an implementation to learn enough to do that.
• Window embedding APIs task has been growing unexpectedly. We need to ensure that we have a handle on the work and are really on track to finishing for moz0.8.

Issues

Escalation

People

• Dave Hyatt is on vacation 1/18 through 1/23.

Highlights

• print range, print selected object are ready to be reviewed for checkin.
• printing predetermined headers and footers is ready to be reviewed for checkin
• Determined that the total page load time on pages with floaters is being impacted by inefficient invalidation of the entire page which results in the entire page being drawn 20+ times. Steve Clark will be looking at making the nsBlockFrame.cpp float logic more intelligent to prevent the full page invalidates and hopefully improve page load time on a number of sites.
• Diagnosed a problem in the revoking of PL_events which is causing random crashers in FrameManager. Dan Mateja has a patch which fixes the problem that he will be checking in soon.

Lowlights

Dogfood

Risk

• Printing the selected Frame bug.. Printing of ART images, and printing offline. These features are dependent on work that Chris Saari is performing.
  

Features

Escalations

Issues

• Rod will need help with "hooking" up the native Mac printer options dialog to take advantage of the new features
• Printing headers and footers. Currently the word "of" as in pages "1 of 2" is not being localized. Need to select a localized font for rendering headers and footers.

People

• Rod Spears tentatively will be on Sabbatical from Mid May thru MidJulyof2001

Priorities for next week

• Don Cone
  • Printing API meeting
  • SMIL meeting
  • Help Rod check in the Printing Selection bugs. Reviews, Mac needsAPI work.
  • Discuss printing issues with Chris Saari. Printing the selected Frame bug.. Printing of ART images, and printing offline. are currently dependent on work he is doing.
• Rod Spears
  • Linux testing
  • Reviews for printing changes/features
  • Getting it checked in
• Waqar Malik
  • 17027 - Eliminate dead code from layout to reduce footprint
  • 29311 - Code to set window position called excessively on GTK
• Eric Pollmann
  • 0.8 bugs (crashers, reduce the size of presstate)
  • 63846 - image inputs don't work on Mac
  • 65099 - linebreaks doubled when text area state saved then restored.
  • 64659 - can't submit form (works for me?)
  • 61557 - location.replace - iframe loads wrong url (finish investigating)
  • Review xforms data model spec
• Kevin McCluskey
  • 49779 - ViewManager clears backbuffer to white before rendering bug triage
  • 62675 - Remove unused code in the view module

Accomplishments

• Don Cone
  • 63549 - nsIImage now supports locking and unlocking.. and will releasethe DIB bits made temporarily.
  • 36796 - Support for Native print settings on the Mac. Partial Checkin.. This was needed for the page range, selection, print options.
  • 65150 - Leaking nsPrintOptions fixed.
  • Working with Chris Saari on some embedding bugs
  • Meetting for the Print API's for printing.
  • SMIL meeting
• Rod Spears
  • Print features
  • Feature Bugs that will be fixed with my checkin:
    • 61075 - No Way To Change Print Range
    • 63426- Printing the selected object needs to be supported
  • Printing Bugs that get fixed because of my checkin:
    • 32128 - Additional information on printing web pages (header/footers)
    • Added additional fields to the XP Printer Dialog on UNIX. You cannow specifiywhether you want to print: All Pages, Page Range, or the Selection.I updatedthe XUL, DTD, and JS files
• Waqar Malik
  • Resolved
    • 64265 Keyboard navigation with CTRL-<arrow keys> crashes build 200
    • 64113 [pp]Submit function is not working when type attribute is as
    • 57589 crashes when opening GAIM link
  • Worked On:
    • 65819 input fields are vertically size of 2 rows17027 Eliminate dead code from layout to reduce footprint
    • 16863 Printing Lock-out (with duplex printing) 44222 Mozilla crashes on nsPrintOptions() during shutdow
    • 45810 Printing non-existent image fetched with Alt-L dumps core
    • 58215 GTK GFX select boxes not updating properly after .add method
    • 54194 Printing applet page is crashing the browser [@ nsPostScript
• Eric Pollmann
  • Fixed
    • bug 55988 (hiddens)
    • 64415 (buttons) [sent for review]
    • Finished fix for 8065 (prevent infinite frame recursion crashes)[verifiedagainst two new testcases, analyzed memory use, sent for review]
    • Finished fix for 42892 - split flushpendingnotifications into contentandcontent+frames cases [sent for review]
  • Investigated
    • found cause of 64412, reassigned to jag, reviewed jag'spatch
  • Bug reviews
    • for edburns (60228, plugins), akkana (56296, diable _new,_blank),jag (64412, double form submit warnings), rods (several), jst (several)
  • Investigated frame targeting problems:
    • 61385: two form submits - second cancels first (-> rpotts)
    • 61187: targetted load cancels <meta http-equiv="refresh"> inwrong frame(-> rpotts)
    • Created bug 65777: loading/targetting tracker bug (rpotts, per jstrequest)
    • 36698 - Analyzed cnn.com testcase - is a dup not a new bug
• Kevin McCluskey
  • Participated in weekly SVG teleconference
  • Checked In:
    • 36849 - synchronous reflow and synchronous paint interact poorly.
  • Worked on:
    • 47549- Entire page redraws as images animate. Identified the location ofthe problem to be BlockFrame reflows. Every incremental reflowis being transformed into a resize reflow and the entire documentis repainted.Did timings to determine how page load times are impactedby the extra repaints.Determined 8-10% of the page load time on CNN.comcan be attributed to theextra repaints.
  • Reviewed:
    • 65243- nsEventQueueImpl::RevokeEvents does not always revoke pendingPL_events

Summary:

• team resolved 36 bugs this week
• continue to work on design strategy for the next version

Individual Status Updates:

• Akkana:
  • 56296: updated the patch and got some reviews (still waiting for sr) for disabling target=_blank.
  • Spent a while tracing the code involved in bug 29346 (disable JS popups) and trying to get the prefs documented; updated the customizing document to show these prefs and several other new prefs.
  • Expanded the fix for bug 64665 (wrap width hardwired to 72 columns) and got it reviewed (not checked in, tree closed).
  • Helped several people on linux install issues.
• Anthony:
  • Worked with aaronl on accessibility bug.
  • Triaged my bug list.
  • Resolved 6 bugs off my list.
  • Worked with Hyatt on some key binding issues, and picked his brain about possible bi-di solutions.
• Beth:
  • worked on sheriff schedule with layout and rendering group
  • continue to work on task planning for the next version
  • triage mozilla0.8 and mozilla0.9
  • attended W3C HTML wg phone conference (minute taker this week) compiled notes and distributed
  • attended several meetings throughout the week
  • debugged incoming bugs, reassigned as appropriate
  • reviewed and commented on W3C Modularization spec
• Charley:
  • Bugs fixed and checked in:
    • 54584 Use global overlay to get last File menu item: "Quit/Exit". Fixed overlay problems with 'Send Page..." menu item.
    • 61996 Enable "Save" command in HTML Source.
    • 65774 Set default for sidebar separator to "hidden"
    • 57649 Fix InsertCharWindow warnings in editor.js
      50328 Click on Table icon when inside cell brings up Table Cell Properties dialog
    • 63971 Add "Save image..." to context popup menu (original fix doronr@naboonline.com; reviewed and improved by me).
  • Bugs worked on but not checked in yet:
    • 56531, 57206 Set 1st suggested word and make "Change" default button and spellcheck dialog layout problems.
    • 60129 Change SetDocumentTitle to be an undoable transaction so changing the title sets the modified flag correctly.
• Daniel:
  • css&fp wg audioconf ; modifications to css3 selectors WD
  • first audioconf with pierre and marc on tuesday about style issues ; will now happen every tuesday
  • fixed 65448 and spent some time looking at the css parser for the :not() css3 pseudo-class
  • finished to integrate comments to my spec
  • technical chat with many people
  • made some technical proposals easily rejected by Joe with arguments nobody could fight against ; *grin*
  • started looking at a way to update the wysiwyg view when the stylesheets in the source view are modified
• Joe:
  • fixed some problems with edit rules initialization exposed by embedding work
  • designed an improved algorithm for the subtree iterator: this should speed up selection
  • continued ws work (ugh)
  • bounced ideas around with daniel about css support
• Kathy:
  • got my development Mac working again (fixed with newest version of Disk First Aid)
  • started looking through e-mail
  • started looking at bugs; investigated a few
  • spent way too much time on the telephone with hr direct trying to resolve way too many problems with payroll, benefits, etc.
• Kin:
  • Code reviews for cmanske, akkana
  • Helped mjudge with linux build support of his trace-malloc changes.
  • Sat with anthonyd to help him look at a couple of bugs he was working on.
  • Did code review with sfraser for find/replace code, and got r= approval from ducarroz and law for the mailnews and xpfe changes. Thanks to sfraser for helping me get it building on the mac. I was going to check it in today (thursday) but the tree was closed until 5pm ... I'll have to check it in next tuesday.
  • Preliminary debugging for some misc. editor bugs.
  • Got the Transaction Manager interfaces XPIDL'ized, and the editor and mailnews code building. (about 45+ editor and mailnews files touched) Just need to tie up some loose ends with some methods Write(), GetUndoString(), and GetRedoString() that I want to remove from nsITransaction.
• Mike:
  • Checked in the trace malloc new files. Have a diff reviewed by kin to not crash build with current flags. With TRACE_MALLOC defined kin is not able to build but that might be from his compiler version.

  
  Brendan has unified diffs and should be reviewing the code and verify that i have caused no harm to the unix issue. I also have some "benign" makefile changes that should only be affected if people build with TRACE_MALLOC defined.

  Still trying to make heads or tails of the huge output i get. unfortunately i cannot run my own debug code because the output file, as waterson says, is about 400 megs. I do not have that much free room in 1 block. I am still messing with that to see what i can do.

• Simon:
  • More work on command handling for embedders.
  • Attended API review meeting to discuss command handling for embedders, had my APIs accepted with revisions.
  • Updated the embedding command handling document.
  • Did a large code review of kin's Find and Replace stuff
  • Corrected and checked in a contributed patch from Jason Bagley (jbagley@artlogic.com) to improve our Apple Events support (can now get the URL of the front window)
  • Implemented the 'inside' parameter for the GetURL AppleEvent, and fixed some other bugs in the Apple Events code.

Lowlights:

• none

Project Issues:

• none

People Issues:

• Anthony out sick 1/18-1/19
• Akkana on vacation 1/19-1/22
• Charley on vacation 1/22-1/26
• Daniel out 1/16 for 1/2 day
• Joe out sick 1/17 & 1/18
• Kathy out on 1/17 -- sick day
• Kin on vacation 1/19 & 1/26
• Simon on vacation 1/26

Summary

• Resolved ~ 38 bugs.
• We have triaged most of our bugs.
• Heikki will be on vacation from Feb 5 to Feb 17
  

Accomplishments

• harishd
  • Triaged more than a 100 bugs - Have divided 20 bugs (approximately) per milestone!
  • Worked on a few of 0.8 bugs ( 58954,48256,43678,65643 - Have fix in hand)
  • Attended Phil's staff meeting.
• joki
  • no status
• jst
  • Major BUG TRIAGE!
  • Worked on random performance and correctness bugs.
  • Finally made some real progress on the DOM IDLC to XPConnect. I created XPIDL versions of our DOM interfaces and checked in those new interfaces, the new interfaces are still WIP but they're getting there. Many thanks for jband his great work on XPIDL.
• peterv
  • Fixed Transformiix makefiles on Windows (Mozilla module and standalone) (bug 46640).
  • Implemented document extension function for the Mozilla module.
  • Merged in various contributed patches that were aging (mainly finishing/correcting the XPath function library).
  • Implemented an XPath processor that returns a node-set posing as a DOM nodelist.
  • Implemented system-property extension function.
  • Implemented function-available extension function (bug 65977).
  • Implemented element-available extension function (bug 65978).
• nisheeth
  • Attended XML Query face to face meeting on Monday and Tuesday.
  • Attended XML Schema formalization meeting on Wednesday.
  • Wrote the conclusion for the web applications document.
• heikki
  • Fixed 64448, comments in internal subset moved to before internal subset in XML documents.
  • Some cleanup of the DOMParser (XMLExtras).
  • Triaged all my bugs; no untargeted bugs left and milestone buckets should be correct up to mozilla0.9.1.
  • Worked on separating content from layout; builds and links on Windows but crashes in startup.
  • Was late from all meetings, somebody donate me a PDA please ;)
    

Plans for next week

• harishd
  • Work on 0.8 bugs.
  • Discuss with Johnny about GetComputedStyle.
• joki
  • Triage all current bugs
  • Get updated on previous two weeks
• jst
  • Do what's left of my bug triaging.
  • Continue working on the DOM IDLC-->XPConnect conversion.
• peterv
  • Implement remaining XSLT extension functions.
  • Continue working on helper class.
• nisheeth
  • Triage bug list
• heikki
  • Finish separation of content from layout.

Escalations

• jst
  • Got a flu which slowed me down some but it's almost gone now.

Highlights

• Chris - Starting an xpapps performance page, adding references to work already done; asked jrgm to post timing page out on mozilla.org. Thanks to Matt & Bill for sending me some performance results/analyses already.
• Bill - investigated load listeners for performance impact, investigated bottoms-up performance items
• Alec - fixed 11 bugs and implemented the history sidebar
• Morse - fixed 4 bugs, producing patch for a host of others.
• Matt - Search issues meeting, looking at whats related and search issues.

Plans

• Continue to look at performance issues
  
• Continue fixing mozilla0.8 bugs and features.

Issues

• Morse - couldnt get PSM to work in the Mozilla build though it works in the commercial build.

People Stuff

• Paul was sick much of the week, hope he's feeling better now.

NSPR Autoconf

• Checked in changes required for AIX to build
• Changed link rules to use EXTRA_LIBS to only link in bare minimum library dependencies. NSPR heavily depends upon shared library dependencies working. Feature matching classic build.
• Sync'd config.sub & config.guess with GNU cvs repo
• Landed changes to enable autoconf to work on win32 (bug 58804)
• Moved from using $(DIST)/bin to $(bindir), $(libdir), etc.
• On win32, force OS_TARGET=WIN95 when building from mozilla. Classic build feature match.

Ports - BeOS

• Bug 65705 - xpinstall/packager under beos is flawed
• Bug 64549 - Implement nsWindow::Enable for BeOS
• Bug 65185 - [BeOS] Bookmarks is not imported correctly
• Bug 65423 - [BeOS] File Dialogs implemented - patch
• Bug 65126 - clicking on the browser does not bring it to the front...
• Bug 65100 - [BeOS: Copy & Paste] implement nsClipboard
• Tested Koehler's mouse cursor changes (bug 63640)

Ports - BSDI

• Set non-monolithic builds to be the default

Misc

• Setup mozilla linux server, myotonic.
• Turned gcc295 nighlty bulids on.
• Bug 55174 - Race condition in creating jars

http://www.mozilla.org/mailnews/status/Mailnews_Status_1_22_01.html

We had some major fixes to 256 color support as well as printing.

We are continuing to fix bugs as we find them. We will be releasing a new build today.

Code is in the process of being reviewed and cleaned up.

Other tasks we talked about last week...

• dprice to work on getting win32 linker to organize hot vs. cold routines. we've made some progress in this area; see bug 65845.
• cls to look into ``magic gcc attributes'' that rth@redhat.com had discussed previously, follow up with wtc and bryner who were trying to post-process the linked .so files and strip exports.
  • cls sent me a private update on this stuff; looks like wtc & bryner gave up; only one of the gcc attributes requires glibc support (``protected''); the others just need a modern linker.
  • IMO there is significant opportunity for .so and runtime footprint reduction if we can strip exports and relocations.
  • Another thing to follow up on is, can we ``pre-rebase'' Linux .so's to avoid rebasing at load time?
• kandrot to get latest Lea allocator wedged into the build, conditioned on an autoconf flag.
• tracy & curt to post ``progress graphs'' to FTP server. I know they've done this. chofmann: could you link their stuff from http://www.mozilla.org/projects/footprint?
• kandrot & thesteve to evaluate ``spikes''; e.g., what objects are live when memory usage peaks on our gross footprint graphs.
• waterson to find and kill object caches. Need to collect new trace-malloc data now that radha's improved session history bloat.

New tasks to track...

• Although we've made enough progress on (1) above to produce linker ordering files, we only saw marginal improvement in Win32 resident set (~3%). Why?
  • waterson trying to figure out what pages are resident after winEmbed has started.
• Revisited the ``RTTI causes bloat'' chestnut with the Gamera project. Turns out adding RTTI information with egcs-1.1.2 doesn't cause much bloat, but it *does* (~20%) with gcc-2.95.2.
  • kandrot dug into this and found a bunch of extra symbols and string in the .so files. (Maybe egcs-1.1.2 just gets RTTI wrong?)
  • The root of the problem here is that Gamera is using exceptions; the gcc exception unwind implementation requires RTTI; for some reason, compiling one module with RTTI entails that all modules must be compiled with RTTI. We should dig into this further to understand the problem.
• pavlov, tor, & saari are whacking imagelib and the image cache. Apparently tor has a hack that will make it so that mozilla doesn't need to keep a copy of the image bits: only the X server needs to maintain them. (Right now, both mozilla and the X server hold on to the image bits.) This could be a huge win: images are the bulk (~20-30%) of the live object data according to trace-malloc.
  • Talked to pavlov yesterday: he was waffly on this subject. Said it was hard. (Waaah! Make it work Pav! :-)

Previous Updates