17:03 <@bz> when I attach with gdb to the firefox process it starts... 17:03 <@bz> gdb can't find libxcom_core.so 17:04 <@bz> or rather ./../../../dist/etc/libxcom_core.so 17:04 < shaver> shouldn't gdb be able to find that? 17:04 < shaver> I guess not if the process chdir'd 17:04 <@bz> shaver: depends on where it's thinking it needs to start looking from... 17:05 <@bz> fwiw, with runtests.pl, the path is 17:05 <@bz> /home/bzbarsky/mozilla/vanilla/obj-firefox/_tests/testing/mochitest//../../../dist/bin/libxpcom_core.so 17:05 <@bz> which is eminently findable even by gdb
17:06 <@bz> fwiw, I just followed the directions in http://developer.mozilla.org/en/docs/Chrome_tests for running a single test using --test-path
This is probably less of a problem with bug 480279, but I haven't actually tried to reproduce, so I'm not sure of the full scope.
I have no idea, since --debugger doesn't work over here.
Though note that this bug was seen on Linux and now I'm on Mac; on Mac I can actually attach with gdb after starting the mochitest process (so this bug is worksforme on mac).