Highlights:
- Space team making good progress with leak fixes, dp's plan
well-scheduled
- Waterson landed RDF-removal from XUL content sink
- Waterson showed XUL content sharing can give up to 4x speedup
for new not-first window
- Travis published webshell design doc/diagram, it's getting
reviewed and implemented
- Pavlov fixed menu dismissal on Linux, xpmenus now dismiss XP
Lowlights:
- No progress on menu positioning or other low-hanging XUL fruit
- Browser dogfood such as session history is not getting porkjockey
help
- Waterson dilemma: make mailnews fast for thousands of messages
vs. make XUL (new window) fast
Highlights
- Resolved 29 bugs in the last week, fixing 20 of them.
See our resolved
bug list for details.
- Eric Vaughan has finished the XPScrollbar integration work.
This required fixing some problems with scrollbars in Ender,
and adding a preference so they could be toggled at runtime.
(evaughan)
Lowlights
Accomplishments
- Mike Pinkerton tracked down some extraneous reflows when
mousing over scrollbars. (pinkerton)
- Chris Saari finshed working on the focus design, and began
implementation. (saari)
- Chris also stopped Mac popups from getting activation. (saari)
- Steve Dagley provided Mac support for the NSPR 3.5 landing
into the mozilla tree. He also reviewed and landed Simon
Fraser's NSPR changes to speed up code fragment loading. (sdagley,
sfraser)
- Stuart Parmenter got menus dimissing on unix. (pavlov)
- Pav also rewrote the Find dialog and fixed some prefs UI
problems.
Priorities
- Fix blocker bugs, improve performance size and speed, polish
the UI, and in general make sure things don't suck for 'beta'.
(all)
- Fix tree scrolling bugs. (hyatt, pinkerton)
- Finish focus work, and then xpmenu bugs. (saari)
- Land XPScrollbar integration. (evaughan)
- Switch over to lightweight scrollports. (evaughan)
- Make dialogs look good. (evaughan)
- Mac work on the revised nsFile implementation. (sdagley)
- Get menu painting not to be drawing images under it randomly
on top. (pavlov)
Decisions
Issues
- There have been more serious regressions lately than should
be tolerated. If a regression prevents you from working
on what you should be, it is just as bad as a build bustage.
Even when working on infrastructure, architecture and performance,
we need to keep the bar high to assure that the builds are usable
every day. Keeping the tree open despite major regressions
just gives the illusion of progress, we need to have the builds
be better every day, with no exceptions. To do that, everyone
is going to have to seriously test their new code in the product
(not just in a test app) before they check in. Saving a
few minutes by omitting this step is a foolish economy which
regularly slows the entire project to a crawl.
People
Accomplishments
- Continued implementation of the network data cache (Bug 8305).
The low-level memory-cache module is now passing its standalone
tests. Upcoming this week: implementing a few stubbed-out
routines in the memory-cache and finishing the cache manager.
[fur]
- Continuing to work on the status bar messages stuff. Limitations
in the proxy event code relating to string copying and proper
cleanup caused the need for a lot of reworking to ensure that
error messages weren't leaked. About to check in. [gagan]
- Continuing to clean up webshell. [rpotts]
- Handed off work on libjar and the jar: protocol to Mitch
Stoltz and Gayatri Bhimaraju. They are in the investigation and
initial implementation phase now. [mstoltz, gayatrib, valeski]
- Working on MakeAbsolute profiling and performance. [valeski]
- Cleaned up ftp implementation. [valeski]
- Figured out a way to get memory statistics on refcounting,
memory leaks and footprint using the existing refcounting mechanism
of xpcom -- the "bloaty" tool. Integrated it with Kipp's memory
work. [warren]
- Fixed javascript: protocol to work again. [dougt]
Issues
- We have a much better handle on how Mac async DNS is going
to work now, and Gordon is in the process of verifying the initial
set of changes to nspr to provide notification from interrupt
handlers. Hoping to have async DNS going this week. [gordon]
- Have an initial implementation of async DNS for Unix, but
it's not integrated yet. [warren]
- URL dispatching -- Checked in initial set of changes for
splitting AsyncOpen out from AsyncRead, although now I'm not
convinced that this is going to help us much. Not much else happened
on this last week, partly because we were waiting on Travis'
web shell redesign documents, and partly because we were just
doing other things. I'm hoping to get going on this full steam
again this week, and will call a meeting shortly. [warren]
Roger Lawrence has
been fixing Spider Monkey and Rhino bugs -
Bugs 14545, 15197, 8976, 15650, 15261, 12168 resolved, one way
or another. Added errors-as-exceptions to Rhino, interpreted &
compiled. Some other Rhino Ecma changes. Fixed up compile failures
for jsfile. Working on 15261, 15604 next.
And Waldemar has
some JavaScript standardization news -
A week ago waldemar, mccabe, and clayton went to Bath, England,
to finish up the ECMA-262 Edition 3 standard. Since then waldemar
made a lot of edits to the final candidate document to reflect
the resolutions from that meeting and fix errors. He sent it out
to the ECMA TC39 working committee. A copy is available on the mozilla.org
site.
John Bandhauer has
been working with Kipp Hickman
to instrument XPConnect code to locate leaks due to reference
count problems. He's wrapping up support for passing arrays and
sized strings to XPCOM interfaces through XPConnect. John has
also updated the XPConnect exception system to be compatible with
Roger's work to expose JavaScript errors as catchable exceptions.
Christine Begle has
landed an XPConnect regression
test suite.
Last week
- 51
bugs fixed (includes both mail/news bugs and porkjockey bugs)
- Team is evenly divided between porkjockey infrastructure and
mail/news bugs for Dogfood and M11
What we're working on right now
- davidmc - Mork memory leaks (15598)
- bienvenu - Memory leaks
- alecf - Converting mail/news components to nsIModule (15161)
- jefft - Non-ASCII chars in subject header (15676), Displaying
attachments inline (12255)
- mscott - Converting to mail/news components to nsIModule (15161),
URL Dispatching (14928)
- sspitzer - Preferences bugs (many), Mac registry crash (15795),
new folder crash (15876)
- hangas - Command updating (12408), moving new skin to mozilla
- putterman - Tree scrolling (10603 and others), nsIModule,
memory leaks
- chuang - Out on sabbatical
- ducarroz - Message sending bugs (14785, 12209, 15235, 15254)
- rhp - Getting back into MIME bugs after sabbatical. Many fixes
last week.
Footprint
- Estimate of compressed total download size on Win32: 1266k
- Uncompressed DLL sides: Win32: 1174k, Mac: 2382k, Linux: 2331k
(down 296k from 9/17!)
Previous Updates
|