2006-12-13 Trunk builds

  • Fixed: 300030 - Refactor intrinsic width computation out of nsIFrame::Reflow (land dbaron's reflow branch).

This is a huge change that David Baron has been working on for about two years. It involved changing 201 files, simplifying many of them: a diff showed 8726 insertions but 18253 deletions, for a net removal of 9000 lines of code. It improved speed on page load tests by 3-5% and fixed over hundred bugs, including:

  • Fixed: 69745 - Auto-width left float containing only nested right float is too wide.
  • Fixed: 129346 - Fieldset renders incorrectly with style="float: left;" or any other shrink-wrap situation.
  • Fixed: 269643 - When clicking link, page redraws with different layout, click is ignored.
  • Fixed: 291126 - Intrinsically sized (shrink-wrap, auto-width) absolutely positioned element containing right float is too wide.

The reflow branch landing fixed the last remaining issues with the Acid 2 test, so Firefox trunk now passes the test.

Other fixes:

  • Fixed: 360493 - Allow users to disable password autofill without disabling password manager, using a hidden pref, for limited protection against certain holes in sites like MySpace.
  • Fixed: 358764 - On Linux, make Backspace do nothing by default. (There is a hidden pref that controls whether backspace goes back one page, scrolls up one page, or does nothing.)
  • Fixed: 347230 - Minimal "Save"-only dialog shown for files with discoverable types (known file extensions) (regression in Firefox 2).
  • Fixed: 341872 - Download manager leaks helperApps.
  • Fixed: 361070 - Implement SVG feBlend and feComposite.
  • Fixed: 351253 - Search button misinterprets modifiers when clicked.
  • Fixed: 339720 - Show warning page before showing about:config.

Fixes for recent regressions:

  • Fixed: 355817 - [Mac] Some Cmd+Shift+letter shortcuts are broken.
  • Fixed: 362812 - [Mac] Don't pass clicks down to controls for ctrl+click.

Trunk regressions:

  • Since Nov 22: 361523 - [Mac] Form borders are too thick (even thicker than before).
  • Since Nov 22: 361576 - [Mac] Underlines sometimes wrong thickness (or missing entirely).
  • Since Nov 22: 361600 - [Mac] Flashing when switching tabs, opening the find bar, etc.
  • Since Sept 29: 355352 - [Mac] While the app is running, if no browser windows are open, clicking on the app in the dock doesn't open a browser window.
  • Since March: 330271 - [Windows] Crashes after trying to check for updates [@msvcr80.dll]. (Fixed only on NSS trunk.)
  • Since Jan 26 (FDL): 324819 - Fixed positioned elements now lag/flicker when scrolling.

Trunk checkins between 2006-12-01 06:00 and 2006-12-13 06:00

Windows builds: Windows nightly (discussion)

Linux builds: Linux nightly

Mac builds: Mac nightly

9 Responses to “2006-12-13 Trunk builds”

  1. Darin Warling Says:

    Great news — I’m really looking forward to v3!

  2. Dylan McCall Says:

    Nice work, David!
    That takes a lot of commitment :)

    It’s great to see FF3 passing the Acid2 test. This release will be quite awesome.

  3. 掌聲鼓勵 « 柏強的城市探險記 Says:

    […] The Burning Edge » Blog Archive » 2006-12-13 Trunk builds * Fixed: 300030 – Refactor intrinsic width computation out of nsIFrame::Reflow (land dbaron’s reflow branch). […]

  4. J. McNair Says:

    Wow, Firefox 3 may officially be the Best Thing Ever(TM). More SVG features, Cairo on by default, Cocoa on Mac, XULRunner, and the ever-lovin’ Reflow Branch. If they can keep (and improve on) the stability and speed, you’re golden. Heck, I’d gladly accept a delay or two for all this.

    Firefox 2 was already Surprisingly Good (though nothing revolutionary). Of course, it did come with the fix for my pet bug (Find in TEXTAREA) which allowed me to kick IE for good. Speedier and more stable than 1.5 even. Mind that I’m on a clean install with fresh profile. I imported bookmarks, etc. after the fact, but that’s standard practice for new Firefox versions for me. I only crashed on abandoned extensions. They should make the “problematic extensions list” much more public, like a default bookmark.

    Basically, thank you for this good blog, and thanks to your Mozilla colleagues for good products.


  5. Laurens Holst Says:

    “so Firefox trunk now passes the test.”

    Wasn’t the nose off by one pixel? I believe that was part of the Acid2 test, too… At least Opera had such a displacement and they fixed it, see: http://weblog.timaltman.com/node/802

    Not that I care :). I’m looking forward to using trunk builds!


  6. Jesse Ruderman Says:

    It is ok for the nose to be off by one pixel. See https://bugzilla.mozilla.org/show_bug.cgi?id=343583#c6 from Ian Hickson, the author of the Acid 2 test.

  7. Artem S. Tashkinov Says:

    Until history backend is moved to SQLite I don’t see any reason for using a trunk. My history is set to 180 days and FF is sluggish as hell (using 100% of CPU) when I try to browse it.

  8. Jesse Ruderman Says:

    Artem, is it worse on trunk than on branch? Or are you using a special build that uses Places?

    You’ll be happy to know that the patch in bug 355738 (history-on-places) just got a patch and has r+. I’m guessing it will be checked in Friday.

  9. Chris Says:

    My nose is off by one pixel too. Breathing works fine. ;-)