status update

maintained by Tim Rowley <tor@cs.brown.edu>

Last Updated Wednesday January 30, 2002

This status update page is updated every week. To get updates and news throughout the week you are invited to check out mozillaZine, a site devoted to Mozilla advocacy.

Previous Updates

Module Updates
XML/DOM/Security
January 29
Submitted by Heikki Toivonen <heikki@netscape.com>

Summary

  • XSLT "rewrite" landed, fixing about 10 bugs.
  • XHTML entities enabled.
  • Lots of minor optimizations landed.
  • Started work on nsbeta1 nominations.

Accomplishments

  • harishd
    • Fixed bug 113349 ( regression ), non-HTML tagnames are stripped out of HTML.
    • Helped heikki with bug 35984, Undeclared entities are ignored when external DTD not found.
    • Worked on P3P related issues.
    • Investigated possible performance improvements in expat driver.
    • Did code reviews.
  • jst
    • Fixed bug 120718 . Made XPCReadableJSStringWrapper not allocate non-shared buffer handles, this caused a malloc for every XPCReadableJSStringWrapper that was accessed in mozilla, causing ~400 allocations at startup.
    • Fixed bug 120901. Eliminated an unnecessary QI call in XPCWrappedNative::GetNative().
    • Fixed bug 120504. Made QI's on DOM elements only call into the binding manager only if the element itself doesn't implement the interface that the element is being QI'ed to.
    • Fixed bug 120899. Changed the NS_WARN_IF_FALSE() macro so that it can't be mis-used in optimized builds in a way that causes debug code to be executed in optimized builds.
    • Fixed bug 121485. Un-nsAutoString:ified 3 perofmance critical nsXULContentSink methods in favor of new string iterator useage.
    • Fixed bug 118933. Sped up the id and name hashtables in nsHTMLDocument by changing them over from nsHashtable/nsStirngKey to pldhash. Also combined the hashes to optimize performance and memory useage.
    • Worked on the W3C DOM Load And Save specification.
    • Started reviewing mstoltz's awesome optimizations in the security manager code.
    • Spent some time working on bug 52334 (display: none; iframes not loading their document).
  • peterv
    • Checked in Transformiix output rewrite (bug 96647). This fixes a number of bugs (including 53944, 54273, 93657, 94921, 107219, 111994, 113215, 113962 and 114414), improves performance and reduces bloat for XSLT transforms.
    • Started working on bug 92377 (Build failure: dom/src/base/nsDOMClassInfo.cpp depends on extensions/xmlextras).
  • nisheeth
    • Did Heikki's focal review
    • Did code reviews for Radha Kulkarni (bug 35340 - wysiwyg: urls - pages created by document.write not added to session history) and Johnny Stenback ( bug 121485 - nsXULContentSink is too nsAutoString happy)
    • Investigated bug 81446 (Reloading page leaks memory) on Windows. Turned out that there were memory spikes after a few successive reloads but, in each case, the memory was freed in the subsequent reload. All in all, not a problem worth spending more time on.
    • Explored whether we can store image data off on the video card on the Mac just like Pav is trying to do on the windows. Read Apple's technote (http://developer.apple.com/technotes/tn/tn1182.html) on GWorlds, objects that can be allocated on Mac graphics hardware memory. Discussed with Simon and Kevin and decided that even though we will get a footprint win if we use GWorlds, we'll run into problems like exhausting application heap space on OS 9 because of multiple handles allocated there for each GWorld. I'm not going to spend more time pursuing this optimization on the Mac.
    • Got together with Akkana to code review a patch from Tanu Mutreja ( bug 85184 - Composer breaks lines at inappropriate positions).
    • Met with Pav to understand his idea about speeding up performance of animated GIFS.
  • mstoltz
    • More improvements and bug fixes for caps performance patch (119646): converted to double hashing, fixed some crashers, re-implemented wildcard policies. Awaiting final reviews
    • Rescheduled lots of bugs - trying to make 0.9.9 list manageable and define 1.0 requirements
  • rayw
    • Bug fixes in SOAP.
    • Started setting up server for testing SOAP services.
    • W3C issues.
  • heikki
    • Code reviews for jst, harishd.
    • Fixed bug 44458, enabled XHTML entities.
    • Fixed bug 35984, made undefined entities always fatal errors. Also fixed a regression when there are errors in an external entity, and made a minor perf fix when loading external entities.
    • Checked in fix for bug 120615, XmlRpc does HTTP PUT when it should do POST. Fix from Axel Hecht.
    • Investigated Bugscape bug 11714.
    • Bug triage.

Plans for next week

  • harishd
    • Work on P3P.
    • Triage bugs.
  • jst
    • Continue working on performance and bloat issues.
    • Continue working on bug 52334.
  • peterv
    • Work on bug 92377 (Build failure: dom/src/base/nsDOMClassInfo.cpp depends on extensions/xmlextras).
    • Fix bug 113351.
    • Work on bug 94036 (Fix xsl:number support).
  • nisheeth
    • Write up Pav's idea on how to improve animated GIF performance and start a discussion on n.p.m.performance.
    • Follow up with Tim Rowley (tor@cs.brown.edu) on Linux patch to store image data in the video card's memory rather than system memory.
    • Investigate the weird discrepancy in memory footprint between "greatest to least" and "least to greatest" traversals of the pages in John Morrison's page loader test tool. Footprint is 10 MB higher for one of these two sequences of page loads even though the set of pages loaded is the same!
    • If I get time, pick off bugs from DHTML performance meta bugs: 21762, 118933, 117436
  • mstoltz
    • Check in performance patch
    • Test and file untargeted bugs
    • Fix 0.9.9 bugs
  • rayw
    • Finish setting up SOAP test server, write more tests, and work on W3C assignments.
  • heikki
    • More bug triage, including nsbeta1 nominations for the team.
    • Knock off another bug or two.
Necko/Imagelib/XPCOM
January 29
Submitted by Gagan Saksena <gagan@netscape.com>

Summary

  • ongoing discussions for priority cache work
  • help from India continues and is much appreciated. Thanks to Vinay and Nivedita for doing excellent job with their tasks. Hopefully Vinay will get his CVS access soon.
  • triaging continues. Numbers are looking a lot better now and hopefully will continue to get better.

Highlights

  • darin
    • fixed bug 121441 "unable to publish via http PUT (nsStorageStream should support nsISeekableStream)"
    • landed partial fix for bug 90004 "proxy reload doesn't work"
    • landed second patch for bug 109179 "replace nsStdEscape with NS_EscapeURL"
    • on going discussions over bug 84186 "8-bit UTF8-capable string type for XPIDL"
    • ongoing design work for priority cache
    • mozilla 1.0 planning / triaged bugs
  • neeti
    • Working on bugs 118884, 118901, 109501, 89327.
    • Triaged bugs.
  • gordon
    • Fix bug 120833 "change default value of nsICacheEntryInfo::expirationTime"
    • Fix bug 119117 "nsDiskCacheMap.h compiler warning - possible data loss"
    • Fix bug 119085 "cache entries corrupted if size of metadata is zero"
    • Fix bug 111049 "add attribute to nsICacheSession to detect device availability"
    • Participated in various "priority" cache requirements/design discussions
  • dougt
    • Reviewed a persistant FTP resume patch.
    • Finished moving callers to new nsIComponentRegistrar. I left a few callers alone. First group is the component loaders. They are still privy to some internal knowledge of how the component manager works. Until we need to define the interface between a loader and the component manager, I feel that we can leave these call sites alone. I also left the plugin code alone. For whatever historical reason, this code reaches behind the component managers back and accesses the component registry directly. This is bad and will break as soon as I change the registry format. I didn't fix this because I was too worried about regressions prior to mozilla 1.0. The last item I did not convert was the python connect code. I have contacted the owner regarding this matter.
    • nsIModule and nsIComponentRegistrar interfaces have been reviewed and will be checked in as soon as the tree opens again. See my posting in the xpcom newsgroup for more information.
    • And Some Confidential stuff...
  • pavlov
    • Checked in fixes for Nivedita. Crash when loading invalid GIF bug 84523. Hooked up the imagelib cache to the memory pressure observers bug 94454.
    • I cleaned up the GIF decoder. Removed some old code. bug 100470
    • Spent a lot of time working on bug 120475 (which turned out to not be my bug). This ended with me creating bug 121673 which now has a patch to handle idle timers.
    • Working with dcone on bug 104999.
  • gagan
    • conf. call with vinay in India.
    • getting a lot of good comments from people using status/reports tool for their focals :)
    • attended meetings with IPv6 guys
    • added a reminder feature when status/reports is past due (more than a week)

Lowlights

  • gagan
    • need to filter out some more meetings from my poor unsuspecting team :)

Plans for next week

  • darin
    • attend fewer meetings ;-)
    • kill off more mozilla 0.9.9 bugs
  • gordon
    • Work on using cache block files for storing cache entry data.
  • pavlov
    • Continue working with Nivedita to fix more bugs.
  • gagan
    • Focals!

Issues

Calendar
January 30
Submitted by Mike Potter <mikep@oeone.com>

Highlights:

  • Fixed many XUL / JS bugs that were causing the calendar to not work properly, or look good, under the latest Mozilla builds. These have now all been fixed, and you can add / edit / delete events with no problem.
  • Important Change: Changed the location of the data file from .oecalendar to "CalendarDataFile.txt", which is now in the user's Mozilla profile directory. Just move the old file to the new location if you've been running the calendar and want to keep your old data.
  • Updated the XPI for both windows and Linux.

Lowlights:

  • We still need help with XUL / JS bugs, and implementing new features.
  • We need help creating a Mac build for the calendar.

Previous Updates