The weekly report for week 27.
printf time! There’s something weird going on
that breaks the XML incrementalism most of the time. Since this
stuff depends on the wall clock, I can’t use gdb to figure this
out.
More detective work figuring out how the sink state is different when incrementalism is broken.
Found the problem. The cause was an uninitialized variable. It worked when allocated from a zerofilled page and failed otherwise. Fixed it.
Made MathML mtr monolithic in the sense tr
is monolithic on the HTML side. It didn’t help with the MathML
problems (see last week).
My new ISP stopped routing packets beyond the first router. I lost access to test cases, communication and MXR/LXR again for a couple of hours.
Blogged about the MathML situation.
Posted questions to the newsgroups.
Restored XSLT and pretty-printing functionality.
Started work towards making opt builds for testing (Mac OS X Universal, Ubuntu x86, Ubuntu x86_64). Takes a surprisingly long time.
Segfaulted on Ubuntu x86_64 on app launch. Pulled a fresh tree the tip.
Updated Developer Tools in order to cross-compile for Intel OS X.
Discovered that the x86_64 build was crashing in my code after all. Interestingly, the Mac OS X PPC build doesn’t crash there. Fixed it.
Published builds.
Noticed that I had been sloppy about handling requests to interrupt the parser. Fixed it.
Mozilla normalizes line breaks twice, which is bogus.
Found a page in the wild that isn’t properly zoomed by the text zoom.
Prepared a test case for the case where the network is not slow but the document is huge. (Using the Web Apps 1.0 spec as a test case again. :-)
Added more parser interrupt opportunities.
Fixed the case where Expat suspends on its own so that the parsing continues later instead of the parser getting blocked forever. (I have no idea what actually happens that makes Expat suspend without the content sink requesting it, but it can happen…)
Attached a patch to the bug.
Tried to figure out how to stop scripts—unsuccessfully yet again. :-(
David Baron advised me not to try to work around the MathML issues by making special cases in the content sink.
Found out that I have broken innerHtml for
XHTML. Planet Intertwingly
crashes.
I’ll be on vacation from Kesäkoodi for two weeks. I’ll continue on July 24.