- Eric Vaughan and Stuart Parmenter got scrolling to blit on
Linux, although a few problems with this need to be fixed before
it can be checked in. (pavlov, evaughan)
- David Hyatt co-hosted a XUL/RDF chat, with Chris Waterson
and Robert Churchill, on MozillaZine. (hyatt)
- Resolved 53 bugs in the last week, fixing 18 of them.
See our resolved
bug list for details.
- We haven't had enough volunteer typists to capture the output
from David Hyatt's fertile brain. He's actually been reduced
to processing the output of a cable-driven electron gun, delivering
DOM-less content known only as "WB" to a phosphor-coated screen.
Please, if you have anything that Dave can help with, that does
not involve typing, let us know. A mind is a terrible thing
- Daniel Matejka fixed nsIEventQueue to not lose NSPR events
when modal windows go away. The corresponding leak remains --
for now. (danm)
- Daniel also got James Toder's Windows window title patch
checked in and credited. (danm)
- Steve Dagley verified more NSPR 3.5 changes (now that
NSPR 3.5 has been moved to the Mozilla tree the NSPR group is
more sensitive to XP changes affecting the Mac so he pre-tests
the landings) (sdagley)
- Steve also fielded some Communicator 4.7 and Mac OS 9 compatability
firedrill investigation and PR. Some part of 4.7 accessed
some forbidden low-memory globals, which causes an exception
on OS9. (sdagley)
- Eric and Pav also figured out a way to remove native widgets.
- Eric also fixed tab view to hide widgets instead of compressing
- Eric worked on cleaning up some xpapps dialogs too.
- Pav also got Unix packaging code up-to-date so .xpi files
can be built on Unix. (pavlov)
- David Hyatt has been working on 'brutal' sharing of content.
- Dave also reviewed the new UI for next week's landing.
- Chris Saari got DOM level focus events and user level focus
interaction much improved on Mac, ie. pretty close to 4.x.
Working on adding the user level stuff to Windows and then Linux
before committing. (saari)
- Danm: 15856, 16551.
*16552 - general unix painting bugs
* 2564 - scrolling large docs on linux does bad drawing
- Evaughan: Linux scrolling, native widget removal.
- Saari: Windows & Linux focus work.
- SDagley: 16202 (to be landed tomorrow and nsIFile implemenation
for the Mac (bug # unknown)
- Hyatt: brutal sharing.
- Finish Windows & Linux focus work and checkin this weekend.
- Kill some XPMenu bugs and manage my bug list. (saari)
- Fast scrolling on Linux. (pavlov, evaughan)
- Get rid of native widgets. (evaughan, pavlov)
- Clean up dialogs. (evaughan)
- Fix typing problems. (pavlov)
- Brutal sharing. (hyatt).
- Mike Pinkerton is on vacation, returning 10/25.
- 41 editor bugs were resolved this past week
- 28 open M11 bugs, 7 are PDT+
- Steve made significant improvements in text control performance
- Akkana worked on IME issues most of the week
- Kin checked in the makefile so spell check will work on Linux
- Charley is making significant progress with the Editor UI
- Simon continues to help improve performance and startup issues
- Kathy is plowing through the key event issues
- Team has completed their goals and objectives for the year.
- Team has 213 total open bug count. Steve and Joe are the most
doomed at this point.
- Didn't get a status from Mike or Joe
- Working on major text control performance improvements for
bug 14727. A 300 text control page now loads in about 2 seconds,
rather than over 1 minute in optimized M10 build. The webshell
redesign will help with the performance (space) problem even
further, so there's less penalty per focus'd control. We can
now consider the space/time tradeoff of having just a single
webshell (soon to be "docshell") per page.
- Nearing final design for the webshell. Hope to have design
finalized Monday, and do a review with jar, hamerly, rickg, kipp,
troy, rickp, hyatt on Tuesday. Result of review will be
to either go with our design, or find a way of scaling back our
design to meet time+risk factors.
- Fixed 16706 -- a blocker in InsertHTML for mail quoting and
- Fixed 15463 (which in turn fixed 16542, 16670, 16645, 16844,
16795) needed to fix the Unix XIM code and re-enable it.
- She also fixed: 16998 (IM blocker, hiding the illegal attribute
on our new <br type="_moz"> tag), 16726 (tracking a change
in doctype handling in the parser), 16441 (entity conversion),
16285 (crash in the copy code on Windows), 15777(a gcc compilation
- Began work on bug 16176 need to annotate line frames with
- Spent some time looking at selection redraw problems that
happen on Linux when scrolling.
- Checked in top level makefile changes to build spellchecker
on Linux by default.
- Working with mjudge on selection issues
- Fixed a slew of bugs this week -- 16923, 12515, 16493, 16494,
- Continuing to work through UI issues: fixed 16868 (H.Line
dialog), 9758 (link dialog), 16691 & 16690 (image properties
- Continued to work on startup performance (see bug 17033)
- Working on pre-loading scheme for Mac libraries, which may
- Wrote some code in Mac GFX to make image drawing better and
to enable 8-bit alpha masks for PNG.
- Assisted Valeski land some changes to FTP
- Worked on key event triaging
- Fixed Mac build bustage for Kipp
- Helped with integration of a net contribution for an Image
Map Editor. Other bugs are affecting the functionality of the
- Helped Kipp with bug 16886
- Triaged editor bug list
- Participated in the Netscape Retention focus group
- International development support by the entire Ender team
when needed and when necessary
- Continue to push on with M11 issues
- Next week:
- Akkana: 15463, 17079
- Charley: checkin UI fixes (6278), finish up image properties
(17098), continue with page properties (14344) and proceed with
- Kin: 16176 and selection issues
- Kathy: update key spec with Akkana; work on imagemap editor
with Charley and others; work on bug 13438
- Simon: Porkjokey and bug 5085
- Steve: webshell rewrite (13374), text control (14727) and
PDT+ bugs (10683 & 15001)
- Joe: IME support and bugs 16432 & 16680
- Simon is continuing to work with Porkjockey effort, editor
tasks are on hold
- Steve is working on webshell rewrite, other editor tasks are
- We are blocked by bug 17015, this needs to be resolved
so we can continue to work on UI issues, we need to get rods up
here to help us work through it.
- Steve will be on vacation the last 2 weeks of November
Ender Team Members:
Pavlov got blit-based scrolling to work, but native widgets
used for iframes, select inputs, text fields and toolboxes are
still a problem. Native widgets in input control are planned to
switch to GFX.
Evaughan: working on a demonstration with no native widgets,
to see how fast this can be. Disadvantage: applets can have window
GFX Scroll bars: still not on by default because of glacial
Linux scrolling, an editor crash double-clicking on URL bar (mistaken
assumption that frame tree is isomorphic to DOM tree) assigned
to Buster, and missing arrows on Mac/Linux. Need someone to run
Quantify on Solaris to find real delays on Linux. Evaughan to
look into speeding things up by remove widget from webshell, if
that is easy.
Need nsiFile work for architectural completeness/correctness.
DougT already loaded with SSL & Webshell work. Travis is talking
more about incremental changes to webshell, rather than a wholesale
rewrite all at once. This work will still take more horsepower,
RPotts is likely candidate to help out.
Vidur: Content notification going well, table bugs getting
resolved. Tables now rearranging themselves as more content comes
in. Some bugs, Vidur is fixing them and hopes to land tomorrow.
In the meantime, he is going through the top 100 sites, will get
a big review tomorrow from Kipp & Troy. Speed is quite good on
Linux (thanks Pav!), not quite as good on Win32. Brendan will
call Troy to see if he has cycles to help out.
Waterson: Brutal sharing: should be done in 2 weeks, with bulk
of the performance gain (content model sharing) in by middle of
Presentation went well, Ref count balancer went over very well,
pizza went over even better.
SmFr has a graph of load time plotted against library size,
by component. Talked with Nisheeth about splitting up layout dll,
since it is way out on both scales. Also has a list of all dlls
loaded, so you can see if any are being loaded that shouldn't
be. No obvious big candidates to be split off to load after startup,
some incremental things like data sources split from RDF.
May be possible to combine some libraries (e.g., XUL into Layout)
to get a speedup. Some things being loaded that could be delayed
(OJI, Wallet, Plugin, Cookies). Need finer-grained analysis using
Quantify, to find out, on a method basis, what is actually being
called at startup, when, and how much space the remainder takes.
May be able to group them based on locality of temporal reference,
or pipeline the loading. Could also combine some with the app
so they load all at once. Some components take much longer to
load than their size would indicate (widgets), why? Appshell,
Widget et. al. also still include native widgets used only by
Viewer, we should split them out to avoid loading them in Apprunner.
Forms code includes duplicate code to deal with native and GFX
widgets, but shipping product will not need any native widget
support. Other components have some duplicate code too.
Need to find out why Layout is so large, is there any way to
reduce it or slice it up to load faster at startup? May be able
to reduce size of binaries by 25-30% on Mac by building with exceptions
turned on, using new tools which may not be ready for prime time
quite yet. VC5 tools produce consistently more compact binaries
than VC6, why? Apple sped things up by combining, then stripping
dead code. Could we do something like that? Could stick everything
in one project, measure what got stripped, just to get the data.
Shaver is working on adding editing to Titledbuttons, using
some XUL popups.
Schedule: Need to get to architectural completeness,
APIs ready (Mozilla alpha?), before getting an interesting feature
set (Moz beta). API review process will run into next year, but
they should be 75% done by alpha.
- New skins are looking great, and planning to land on Tuesday
- Win32 download footprint (compressed, all code and resource):
- Compose window improving
- UI design meetings. Last week:
- Sent/FCC folder behavior
- Attachments as link
- UI design meeting next week:
- merging Account Manager back into main prefs
- mail/news menu structure
- What we're working on
- davidmc - logging tools for Mork, sorting, 10837
4.x AB import
- bienvenu - 16839
IMAP flag synchronization, 12491
Filtered messages not being shown in folder
- alecf - Picking up tree widget bugs from hyatt and putterman
- jefft - 16393
XP problems with text attachments, 16338
POP attachments not downloading
- rhp - 16703
Crashes in InsertAsQuotation, 16367
Attachment as link
- mscott - 16927
Images in messages not appearing in replies, 14928
- sspitzer - 17078
Assert/crash in subscribe dialog, 16341
More subscribe bugs
- hangas - 15127
Command updating and dispatching, 16395
Convert to new skin
- putterman - 12846
Make message threading work, 16545
make mail check notification work
- ducarroz - 17025
To field is empty in reply, 14112
Japanese recipients not converted to ISO 2022
that the previous editions missed, including exceptions, try/catch
blocks and regular expressions. Check it out at mozilla.org. (PDF format)
News from Roger Lawrence:
Resolved 14060, 6704, 12951, 15593, 15197. Other sundry bugs
in Rhino, Monkey. Working on proper ECMA handling for all undefined
or NULL args to functions. Changing test suite to match ECMA 3
specification. (previous versions is tagged)
Next week - more of the same, plus JS1.5 delta doc stuff.
SPARC build failures still unresolved.
Lots of news from John Bandhauer:
- I continued fixing little leaks and one corruption bug in xpconnect.
- I implemented nsIXPCNativeCallContext as a system for native
code called from JS to get at underlying JS args etc.
- I fixed nsIXPCSecurityManager filtering problems in xpconnect
to deal with problems that norris is running into.
- I spent some time thinking about JS debugging in mozilla issues
so that I can point the UWO team in the right direction and help
- I'm looking for help in ownership of issues and code regarding
xptcall on 'nix platforms. Issues regarding binary incompatibility
of xpcom components built with various compiler flags and libraries
are starting to surface again. Here is a
posting I did about the issues some time back... Besides the
coordination of 'nix efforts issues, I'd love to find a qualified
owner for the (functional but suboptimal) Linux version of xptcall.
- I started design work on xpconnect refactoring needed to fix
a number of bugs on my list. This requires a lot of juggling.
I will try to sort it all out into a presentable form and get
some design review.
- bugs (all relying on xpconnect refactoring): 13419 13429 13425