2008-10-26 Trunk builds


  • Fixed: 412486 - [Mac] Widget/event-detection support for multi-touch trackpad gestures.
  • Fixed: 456520 - [Mac] Multi-touch gesture support for Firefox. (Using the trackpads on some newer Mac laptops, you can now swipe with three fingers to navigate through history or scroll to the top/bottom of the page. You can also pinch to zoom and twist to switch tabs. Read more on Mardak's blog.)
  • Fixed: 396816 - Location bar should be self-describing: "Search Bookmarks and History".
  • Fixed: 400398 - Remove the throbber from the menu bar / nav toolbar default set.
  • Fixed: 449522 - Context menu for HTML5 <video> elements.
  • Fixed: Add hidden pref layout.css.visited_links_enabled for users who want to prevent sites from being able to take advantage of bug 147777.
  • Fixed: 398928 - [Mac] Allow moving a unified window by dragging its toolbar.

Fixes for recent regressions:

  • Fixed: 459323 - Drag and Drop in input controls is broken.

Trunk regressions:

  • Since ~Sept 26: 457187 - 'New Tab' button should be customizable.

mozilla-central pushlog for 2008-10-17 04:00 to 2008-10-26 04:00

Windows builds: Windows nightly (discussion)

Mac builds: Mac nightly

Linux builds: Linux nightly

5 Responses to “2008-10-26 Trunk builds”

  1. Scott Gifford Says:

    I just installed this nightly, and it takes a long time to start up. A little experimenting with strace and lsof shows that it’s busy writing 2GB worth of zeroes to a temporary file with a name like /tmp/jemalloc.YyvBPf. After it finishes, it closes the file so it is deleted, and seems to run normally. Yikes! Any idea what’s going on here?



  2. Jesse Ruderman Says:

    Scott, a Google search led me to https://bugzilla.mozilla.org/show_bug.cgi?id=447710, which in turn led me to https://bugzilla.mozilla.org/show_bug.cgi?id=448840. Neither bug mentions slow startup or writing pages and pages of zeroes, though. The intent seems to be to reserve disk space to prevent overcommitment.

    Are you using an unusual file system? How much does MALLOC_OPTIONS=O help?

    I think you should contact Jason Evans and/or file a new bug depending on bug 447710 in Core:jemalloc.

  3. Scott Gifford Says:

    The problem goes away when I set MALLOC_OPTIONS=O. What does that do, turn on the old allocator? I’ll file a bug shortly, thanks for the help!

  4. Scott Gifford Says:

    OK, reported bug 461782. Thanks again!

  5. Jesse Ruderman Says:

    According to a comment in bug 448840, MALLOC_OPTIONS=O “enables overcommit”, whatever that means. It’s still using jemalloc.