On Mon, Mar 12, 2012 at 10:46:28PM -0400, Brian Callahan wrote: > Hi ports -- > > So in my attempt to have once again a fully working Webkit for > Loongson, I have run into a snag. I have two patches at the bottom > of this email that, when applied, will allow webkit-1.6.3 to build > and run on loongson. > > So far with these patches: > *The SunSpider benchmark runs and completes. > *On webkit-gtk3-1.6.3, the Google search as you type feature works. > *Some JavaScript works. > > However, a lot of JavaScript does not work and crashes webkit and > the browser. I've discovered that the quickest way to get a crash is > to go to gmail.com and attempt to log in. > > Here are two gdb backtraces from logging into gmail.com (both with > the gtk2 version of webkit), the first compiled normally as per the > Makefile, the second compiled with -O0: > http://devio.us/~bcallah/gdb-surf.txt > http://devio.us/~bcallah/gdb-webkit.txt > > I was hoping that someone could look through this and see what I've > missed, and hopefully get webkit fully working once again. If > there's something else you'd like me to include, please let me know. > > Everything is being compiled on a Yeeloong, running -current with a > GENERIC kernel. > > The "ideal" solution to this issue would be to enable the JIT on > Loongson, but our assembler is missing the .reloc pseudo-op, which > is necessary to compile the JIT for MIPS.
Probably not the reply you're looking for, but you should report all that (patches and backtraces) to bugs.webkit.org. That wont make sure someone looks at it, but at least it's tracked somewhere.. Fwiw, webkit is badly broken too on sparc64, and crashes quite often on macppc too but with a different backtrace (crash in gmp, see https://bugs.webkit.org/show_bug.cgi?id=79567). Oh and of course, when posting backtraces, make sure webkit is compiled with debug information (DEBUG=-g in the build env 'should' do it) In the meantime i'll commit your patches, but that's all i can do.. Landry
