I am happy to announce that NSS 3.1 Beta 1 has been released,
including a new implementation of the RSA algorithm. This
release provides, for the first time, a complete open-source
implementation of the Netscape crypto libraries and will be
used in a future version of Personal Security Manager for
The release notes can be found here.
The final NSS 3.1 release is scheduled for Oct. 13, 2000.
Please help us test NSS and port it to your favorite platforms.
- with help from samir, dveditz and mscott, I [Seth] fixed bug 51267
. This was the much dreaded "intermittent failure to load CSS files from
JAR packages". (I stole the bug from warren, who is most-doomed.)
- also, mscott helped joe hewitt fix some nasty build bustage late last night.
- Thanks to Andrew Wooldridge
for adding a focus border to the 3 pane so we now know which pane has focus.
Thanks to dbaron for helping with the 1MB per compose
window memory leak, which is now fixed, and thanks to Hyatt for fixing the
xbl event handler bloat on text elements.
Thanks to Conrad Carlen for providing help with DirectoryServiceProvider
for StandAlone apps. Also, thanks to Ben for helping with some nitty-gritty
chrome rules and dynamic overlays.
- Infrastructure regressions in general getting very expensive
- Lost a lot of time Wednesday recovering from the progid landing
which wasn't pre-flighted on the mail application.
- Lost a lot of time Thursday night trying to fix a crash on startup
on all 3 platforms that got checked in just before the tree closure deadline.
Not having people who checked in the offending code available and willing
to help track down these kinds of errors is a continuing problem. Basic
smoketesting such as running the app would have caught this and saved us
lots of time.
In short, poor landings and lack of responsiveness by
people checking in and not helping to track down problems they introduced
continues to cost us time!! These incidences could easily have been prevented
and the fact that they happened the day of and the day before a major tree
closure is all the more upsetting....
- Another happy customer: "I have been extremely frustrated this week just getting
a commercial build that I can run successfully. I still don't have a commercial
build that shows the overlay stuff even after going through all the hack
suggestions posted and noted on tinderbox."
- And another... "another week with a lot of regressions and problems with the build."
- And another... "Tree closures and unstable builds towards the end
(of regular checkin deadlines) took a hit on planned checkins."
- And another... "Worked on and verified some regression
bugs caused by the build last week."
- Still high nsbeta3 incoming rate
- Need to find an owner for #51571, since it doesn't look like anything we
Vacations & Sabbaticals!
|Date||Open nsBeta3+ P1/2||Open nsBeta3+ P3/4/5||New, 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+)|
(26 were P1/2)
(29 were P1/2)
(16 were P1/2)
(20 were P1/2)
- Bugs 48981 (standalone app support, shrimp being one of the consumers)
and 49088 (dynamic overlays for activation) are fixed and tested. They
are under super-review now.
Ray landed his changes to the format of progIDs (now called "contract IDs")
-- the single biggest checkin in history! Checking in 1000+ files brought
down the build machines due to a flood of email. But we're up and running
with them now.
Chris updated his machine to a 750 mhz one -- he's going to be twice as
Jar packaging is taking longer than we thought to land. We had some problems
with the Shrimp installer that caused us to back out the jar.mn changes.
Now we're in the process of reinstating them with improvements that work
Fix RDF and XUL template arena sizes based on some analysis
of the PLArena stats. Checked in, and reduces per-window footprint by 300K.
Profile to confirm that dogfood+ bug 50634 ("CNN locks up") is related
to async reflow. To nisheeth now.
Fixed 46686, crash when booking same URL twice.
Fixed 51813 (can't flag messages sometimes), a regression that was introduced
when hyatt and I hacked stuff for menu performance in nsbeta2.
Really fixed 46134, support >1 value per attribute in XUL templates.
Worked on memory flusher stuff, see bug 44352.
Additional work on nsbeta3+ bug 50071 - XHTML through the XML codepath.
Investigated responsiveness problem on general page load. Tracked it down
to patch for bug 36849 (which was subsequently backed out). Discussed event
prioritization on Linux with Pavlov. Additional discussions are currently
taking place on porkjockeys.
Worked with email@example.com
to come up with an alternate codepath for editor saving, cut-and-paste
and mail quoting that doesn't involve XIF, since XIF introduces considerably
bloat and performance cost.
Ran Footprint/Performance meeting. Notes posted to porkjockeys.
Several internal meetings on web applications.
This week was preoccupied with landing the progid changes, keeping the
changes merged until I got permission to land it, and fixing a few problems
that the change caused. I also spent a lot of time talking to people
about technical, strategic, logistical issues.
I missed the DOM teleconference due to being busy with progids.
The DOM working group voted this week (via e-mail) to finally send the
DOM spec down the path to become a recommendation.
Spent considerable time discussing the component manager and service manager
API changes, which have not been made.
I have neglected serious DOM editor work that, now, must be complete before
Wednesday, which I will have to make my highest priority for the first
part of next week. The work of the DOM WG is heavily heading into
level 3. It is also critical that I get David Hyatt to review and
compare the API next week before the DOM face-to-face meeting in Boston
in a week and a half.
I will try to work towards codifying the API changes that we have worked
out thus far at the end of the week, but I really feel that working out
a post-6.0 strategy is best. There is, for example, enough other
things in the XPCOM APIs that need to be cleaned up that I feel it might
be appropriate to create a compatibility XPCOM that only loads in the presense
of old clients. If we cannot work out a reasonable migration strategy
for this type of thing, we are doomed before wed get started.
I am also trying after-the-fact to verify that external projects such as
OJI and plugins that might have been affected by the progid changes were
Transitioning XPCOM work to Edward Kandrot.
Worked on ECMAScript Edition 4 spec and tools.
Worked on memory log visualization tools.
Attended Vidur's meeting.
Played virtual Warren for one more day.
Worked on ECMAScript Edition 4 specification.
Met with Mike Cowlishaw from IBM to discuss ECMAScript arithmetic issues.
Began planning for England TC39 meeting.
Did my bit to chase, report, and help fix the current crop of bugs uncovered
Did more prep work on bug 46707 to get us using typelibs stored in zip
files to reduce memory footprint. This looks like only a 50k-100k win.
Fixed bug 13125 to allow the use of xpconnect wrapped native objects as
There we a number of people clamoring for this.
Spent time chasing Win98 specific bugs - without a great deal of success.
Spent time brainstorming with vidur, beard, and rayw on new web application
Working on jar packaging.
Checked in jar packaging for unix, but disabled it until jar file restructuring
Worked on adding the + feature to jar.mn files to allow overrides.
API changes to nsSlidingSubstring that Vidur needs for his parser work
made flat strings use .get() rather than implicit operators; still need
to outlaw the operators, though
helped other people; reviewed, tested, and/or checked-in their patches
helped Pav with C/C++ syntax
reviewed jdunn's patch
modified Mac projects on shaver's behalf
modified Mac projects on edburn's behalf, and helped him understand the
best order for checking files into the tree
sheriffed on Wednesday
helped Simon Fraser hack perl for build scripts
macdev meeting (1.5 hours)
mozilla.org staff meeting (2 hours)
managed to install Yellow
Dog Linux on my PowerBook; got X working ... soon I hope to have working
Airport support so I can actually pull a tree and build mozilla
Implemented and checked in a fix for part one of bug #46709, which deals
with Component Manager startup time and bloat.
Learned all about build breaks and why one needs one of each platform to
Found and filed bug #51944, which deals with bloat in CSS's CSSStyleRuleImpl
and dup'ing of data.
Fixed bug #51944, which removed 200k of bloat.
Changed the way prototype attributes for XUL are stored, reducing bloat
Set up the Win machine for which I had been waiting.
Sat in on the Mac Dev meeting to find out some of the Mac performance issues.
bruce AT puremagic.com
Fixed #50876, the nasty Solaris bustage.
Filed bugs on:
UMRs: 50849, 50851, 50864, 50871 (dup), 50889.
MLKs: 50614, 50856, 50858 50861, 50862, 50863, 50866, 50878, 50884, 508886,
50887 (dup), 50888 (dup).
ABRs: 50498, 50499, 50613.
FMW (free memory write): 50867
The XPToolkit team resolved 64 bugs in the last week, fixing 40 of these, including 24 '+'. For details,
- We have 13 + bugs remaining.
Mike Pinkerton (pinkerton)
- Use filtering based on file type if we fail on the IC check in file picker dialog (50312, 51830)
- fix regression of mac check menus (51685)
- spinning watch cursor on mac when gecko locks up UI (52108)
- fixed DnD getting in the way of listening to DOM mouse events (43258, 52541)
- fixing access key conflicts with menus (40071)
Chris Saari (saari)
- fixed a few nsbeta3+ bugs and hunted regressions.
- Focus is better than it was at the beginning of the week, with the exception of a few new issues on the Mac which I'm working on (particularly, initial focus on new windows got broke). I'll need danm's help with those next week, as it looks like dialogs are forwarding the deactivate event from their parent to the dialog (which we really don't want). Once that is fixed I think initial focus with textfields (like Open Location) should start working on MacOS too. This may well be the reason that dialogs have never worked correctly with focus memory!
Daniel Matejka (danm)
- (on vacation nearly all of last week)
Eric Vaughan (evaughan)
- Fixed numerous bugs: 49408, 52124, 51894, 30548, 37836, 40951
Stuart Parmenter (pavlov)
- lots of reviews some work on 51603
David Hyatt (hyatt)
- fixed xbl security bugs
- fixed tree widget bugs
Dan Rosen (dr)
- resolved 5026Ill-formed browserBindings.xul crashes mozilla
- helped with 39957-family of bugs as much as possible
- fix in hand for bugscape 2324, 1896, possibly 1292 (should make nscp people's lives much easier in coming weeks).
Brian Ryner (bryner)
- 51952 (disable NS_BOX_ASSERTION in release builds)
- 51903 (assertions switching pref panels)
- checked in contributor patches for 52054 (incorrect gtk system color), 41330 (shift-up/down in trees)
- made us not dependent on building extensions/psm-glue (18352 -
Crash when entering the subject of the compose window
Help menu needs to support submenus
Tooltips interfere with accessing context menus
Mail Composition: Cursor not present
editor does not regain focus after dialog is dismissed.
style buttons missing
- Help out others with their nsbeta3+ bugs
- Helping on 2 beta3+ bugs one from saari and on from pinkerton.
- Hope to be available to help on others again very soon.
New window performance
Strange little windows are dangerous for your health
Crash exiting browser
Sidebar Links opening in Composer window.
Unable to use the app after creating a News Account and gett
- 52217 (need help tracking down where this is happening)
linux, mac?: Net2Phone shouldn't be in the toolbar customiza
- help investigate bugscape 2321
- 47395 may become high priority again, as 24645/52359 is currently infeasible
- 52216 (tree crash on item removal)
- 51833 (disappearing folders in ftp)
New window performance
New features still landing by
way of mozilla.org can cause regressions in Seamonkey, although they are
not necessarily needed in Seamonkey. One such feature landed Friday, after
testing on only one platform, and broke all Mac sub-menus, contributing
to long tree closure, and sucking up significant Netscape engineering time.
I can't help but wonder how many other regressions we're seeing have a
similar cause. I know the reasons for staying on the trunk this long, but
at some point the risk becomes too much relative to the time remaining.
We need to branch ASAP.
We currently have 12 nsbeta3+
bugs with a fix in hand for 4 of them.
We currently have 11 beta 3 nominated bugs.
The XML/DOM team resolved 38
bugs last week.
Tom Pixley got rid of 5 nsbeta3+ bugs and reduced his nsbeta3+ bug
count to 0! Great work, Tom!
Peter van der Beken (firstname.lastname@example.org) has enabled XSLT
hooks by default in Mozilla. This means that people can build the
Transformiix XSLT processor, add it to the components directory, and start
playing with XSLT transforms in Mozilla.
Tom Pixley is in Mountain View this week, from Tuesday, September
12 through September 19. He will be using the machine in the cube
two spaces away from Clayton. Stop by and say hi!
Nisheeth Ranjan will be away to New York to attend the XML Query
face to face meeting from Wednesday, September 20 through Friday, September
Accomplishments of last week
- Harish Dhurvasula (harishd)
Checked in fixes for a couple of beta3+ bugs
Worked on bug 49115 ( Will be applying the patch to a good tree! )
Worked on bug 50213.
Triaged my bug list - Current Status: 2 beta3+, 1 nominated, 3 untargeted
- Heikki Toivonen (heikki)
Sent a message to the XML Xlinking WG about Netscape's position with XLink,
XML Base and XPointer.
Fixed Select All case of bug 46085.
Helped joki with bug 47022.
Working on 47319 (TOC button not working always). Currently no ideas what
is going wrong, event gets fired and it looks like it gets handled (even
though JS is not executed).
I have a partial fix for 50478 (events bubble but report bubbles=false).
We have wrong logic in GetBubbles method. The bigger problem is that just
about every element bubbles in Mozilla. This is P3 right now so the big
problem probably won't be attacked.
Working on 52585 (images load event bubbles which can confuse document
load event handlers). I have a fix but it is still buggy.
Learned about our XML & DOM test suites etc.
- Tom Pixley (joki)
Fully fixed 3 of the 5 nsbeta3+ bugs by the Sept 14th tree freeze
Fully fixed 1 of the others except for what seems to be an edge case or
simply a very complex test case. I marked the initial bug fixed and
opened a new bug for the edge case
Partially fixed the final bug which was a 4x compat bug. More or
less 4x lets you cancel text input either by cancelling the keydown or
keypress event. I made the keypress part of that work in mozilla
but the keydown is much harder and won't happen now. By fixing keypress
we now have a workaround though. I annotated the bug with the new status
and workaround, removed the nsbeta3, and futured it.
- Johnny Stenback (jst)
Fixed nsbeta3+ bug 51562.
Made nsXMLContentSink use a nsCOMPtr to make it not leak nsXMLProcessingInstructions.
Fixed nsbeta3+ bug 51034.
element.getElementsByTagName() was throwing a NS_NOINTERFACE exception
due to the wrong IID passed to a QI call (this also caused a nsContentList
Fixed nsbeta3+ bug 52572.
Updated the DOM intefaces in mozilla to match the current DOM Level 2 PR.
Fixed nsbeta3+ bug 48161.
Fixed getAttribute() on a XUL element to conform to the DOM Level 2 spec.
Checked in part of the fix for nsbeta3+ bug 42321.
Avoid running JS_GC() many times per window open/close, once per window
open is enough. Patch mostly done by email@example.com.
- Nisheeth Ranjan (nisheeth)
Triaged my untargeted bug list.
Fixed nsbeta3+ bug 49579.
Attended a two hour long XML Query teleconference on Wednesday.
Met with Chris Petersen on Friday to understand the current state of XML
testing. Heikki Toivonen attended the meeting.
Met with Gerardo Kraveternik on Friday to understand the current state
of DOM testing. Gerardo gave a demo of the DOM test harness.
Bob Clary, an independent programmer who has developed a JS test harness
for DOM 1/2 testing teleconferenced in and talked about the architecture
of his test harness. Heikki Toivonen and Eric Krock also attended
Attended a day long class on Project Management on Thursday.
Tested MOZ_XSL enabled builds on Linux and Windows and provided feedback
to Peter van der Beken.
Plans for this week
- Harish Dhurvasula (harishd)
Fix my remaining beta3+ bugs
Will try to trace down performace hot spots
Fix leaks if any
- Heikki Toivonen (heikki)
Fix remaining nsbeta3+ bugs
- Tom Pixley (joki)
I have at least one 'egk' bug (on Eric Krock's must fix bug list) to work
- Johnny Stenback (jst)
Continue working on my remaining nsbeta3+ bugs
- Nisheeth Ranjan (nisheeth)
Work on remaining nsbeta3+ bugs.
Attend the XML Query face to face meeting in New York on Thursday and Friday.
nsbeta3+ still pouring in...
regressions still high.
Dogfood and Risks
Worked on cache eviction bugs with dp
Fixed bugs 48292, 48401, 42606
fixed: 13718*, 50778*, 51249*
duplicates: 36785*, 42891, 52079, 52088, 52227
WorksForMe/invalid: 50281, 52785
Reassign: 46871, 52074
current focus: #41268, caching animations.
#47098, throbber MLK
Continuing to triage bugs for the whole team.
Set up a hire status page to allow for better tracking of the process of
Worked fixes for 2 top necko crashers. Haven't check them in as yet.
SSL proxy connections seem to have regressed pretty bad.
No status from ruslan and gordon this week.
As requested by firstname.lastname@example.org, here are the
tree rules Netscape is using as we finish up the Netscape 6 (aka Seamonkey)
Here's a summary of the tree rules, as presented at the 9/13 all-hands meeting
and the 9/12 seamonkey-leads. We're making two process-wise changes:
Here's a timeline of what kind of work we expect to do between now and RTM. The priorities shown are defined using the PDT priority list:
- Fix the right bugs. To make sure we're fixing the bugs which
really need to be fixed (and not the bugs we can live with) PDT is doing
a second-level review on triage approvals made by the eng+PM triage teams.
Triage teams do [nsbeta3+] and PDT does [nsbeta3++], beginning 9/21.
- Fix the right bugs the right way. In an effort to reduce regressions
at this critical time, we're working with mozilla.org to ask engineering
leaders to do more in-depth code reviews. This system will be described
in more detail by brendan, but it's going to use the mailing list (and newsgroup
mirror) email@example.com. We'll start this as soon as mozilla.org has the reviewers ready.
- Criteria generally applicable across the product.
- Some engineers may not have any P1 bugs -- your only bug may be a P3.
- We may not get all bugs in of any of these categories fixed, but it's
a way to compare one bug with another.
- The triage down to 3.5 per engineer (~300 across the project) must
be done by Friday.
- High frequency crash (e.g. top 100)
- Serious data loss
- No workaround
- Major contractual requirement
- High profile partner
- Legal correctness
- High profile backwards compatibility
- High profile functionality
- Obscure work around exists
- Obscure crashers
- Generic standards
- Easily recoverable data loss (e.g. form fill, preference)
- Obscure data loss
- Low profile functionality
- Obvious workaround
- 9/14 - 9/21: Only P1 and P2 bugs can be fixed, with a code review from firstname.lastname@example.org. Managers are minusing bugs rated P3, P4, and P5, to be complete by 9/15, which leaves only P1 and P2 nsbeta3+ bugs open.
- 9/21 - 9/28: Only B3 pull-it-off-the-wire bugs can be fixed
on the B3 branch. PDT doing approval of [nsbeta3+] bugs using [nsbeta3++]
notation. Code reviews from email@example.com
- 9/28 - 10/15: Only P1 bugs can be fixed for RTM, with code review
from firstname.lastname@example.org. PDT doing approval of [rtm+] bugs using
[rtm++] notation. Triage teams must give rtm nominations [rtm+] in order
to get an [rtm++] from PDT.
- 10/15 until it's done: Bake time. Only pull-it-off-the-wire bugs can be fixed, with code review from email@example.com. PDT still doing approval using [rtm++] notation.
Anyone who has difficulty making their work fit in this framework should email PDT or come see us at 4pm in Star Trek (3pm on Fridays)
One last reminder: UI freeze for localization is 9/14 at midnight. Any changes which impact l10n after that must be cleared with the l10n group.
The NSS team met yesterday and decided on the
following 3.1 schedule.
- Fri 9/29
- Checkin cutoff. Code freeze starts.
(Note: we are not planning to branch
for 3.1 at this point.)
- Fri 10/6
- Release candidate.
- Fri 10/13
- Final release (RTM). Tip re-opens
for new development.
We are continuing to make nightly builds available at
Major code checkins recently include the splash screen, as well as fixes to
menus and general nsWindow work.
Our primary focus right now is keyboard/accelerator keys and then we will
move on to font work.
We are merging the branch up to the tip with hopes of a carpool after the
This new roadmap was updated the other day, but it felt "draft-y" and
I didn't announce it while we were getting the code reviewer cohort tuned
up (now splashed -- see also the reviewers
document and the mozilla.reviewers
The new roadmap does not give airtight proofs or use the precise language
of Mathematics. But it addresses some important points raised in
the thread to which this message is a followup, and it defines a rough
schedule and process that we'll be using to manage the Mozilla project,
led by firstname.lastname@example.org.
Your comments are welcome at any time, note followup-to: setting (or
feel free to mail me). I promise to put frequently asked questions
in the roadmap-faq
this time! The drivers will be meeting virtually and coming up with
some definitions for mozilla 1.0 and beyond, and those will show up in
roadmap, or at least linked from roadmap.