Started going through my untargeted bug list. Knocked off a few easy ones, including 59906, 70005, and 70699. Found a few more that seem to have magically started working, including 56733 and 58784.

Hyatt showed me how outliner worked.

Checked in a fix for 70005, a lost root that was sending the garbage collector off into the weeds. Worked up a potential fix for a crash I was seeing in the JS garbage collector after opening the editor on www.webmonkey.com several times.

rfg compiled mozilla into a Big Ball of Mud (one executable), so I went ahead and compared his size numbers to those of the individual .so files. It turns out that his numbers are about fifteen times smaller than the sum of the .so sizes. Smells fishy, especially since The Big Ball of Mud is smaller than some of the individual .so's. Think he might need to include the --whole-archive flag when linking.

Spent some more time looking into 71141, and it turns out that we're actually creating two script objects for a single DOM element here. (The XBL onattach handler is causing some unexpected re-entrancy.)

Consolidated data about frankengcc, and published a page on mozilla.org about rfg's gcc hackery.

Worked out a real fix for 71141.

Spent the last few days working on the RDF outliner. It's now at a point where the only major piece of missing functionality is sorting. It may need a bit of performance work, too. Along the way, I cleaned up the old ugly ``element delegates to document which trawls through all the builders'' to create lazy content. Since the XUL element is knows if it's lazy, and it'll only ever be lazy if a XUL template builder made it, it's safe to just walk up the parent chain looking for an element with a builder attached to it. (Also, I made it so that this ``attachment'' is maintained in the document using a hashtable rather than taking up space in each element's slots.) Finally, I exploded the template builder into several different files (pretty much one per class) rather than trying to pack it all into one big file.

Cleaned up another XUL-without-JS crasher that Georgi found.

Got ``natural order'' sorting sorta limping along over the weekend. Bugged hyatt and ben for review. Bit of help to pierre getting a Linux build with his style changes. Spent most of the day getting rid of eight byte overhead per object in nsFixedSizeAllocator, bug 68213. Ended up being a rather large jihad; put together a patch, soliciting review.

Updated the fixed size allocator patch with feedback from scc, made sure it builds on Mac. Looks to save about 50 to 60KB in live bloat going to the first window. There are about 30,000 allocations from the allocator on startup, mostly from the parser, so I'm not sure how those are affected (but it's gotstabe good).

Came up with a patch for another block-in-inline bug, this one having to do with reframe.

Worked up a patch for an image map bug by making code that grovels over the content model a bit more tolerant.

Tinkered with the layout regression tests, trying to get rid of all the false positives.