The abort() call in AvmAssertFail could be replaced by the components in JS_Assert. Currently abort() calls cause testcase reduction issues (in Lithium) in Windows when a LIR assertion is hit. http://mxr.mozilla.org/mozilla-central/source/js/src/nanojit/avmplus.cpp#65 http://mxr.mozilla.org/mozilla-central/source/js/src/jsutil.cpp#67
We own that file, so we can patch it at will. Go for it :)
Created attachment 450039 [details] [diff] [review] take one
Created attachment 450040 [details] [diff] [review] ignore whitespace nits
Created attachment 450042 [details] [diff] [review] side patch Does this side patch make sense? I'd thought to synchronize the "Assertion failed" vs "Assertion failure" message, but this doesn't seem important except from a cosmetic point-of-view, and I'm not sure if there was any reason for calling it the former rather than the latter.
Comment on attachment 450042 [details] [diff] [review] side patch nanojit.h is shared with adobe. The patch is fine but there is a certain process to apply it (commit to nanojit-central). njn knows the drill.
Comment on attachment 450040 [details] [diff] [review] ignore whitespace nits Looks good.
(In reply to comment #6) > (From update of attachment 450040 [details] [diff] [review]) > Looks good. http://hg.mozilla.org/tracemonkey/rev/2ae7cb9510b3
I pushed these for Gary: http://hg.mozilla.org/projects/nanojit-central/rev/c7e84fb8ce71 http://hg.mozilla.org/projects/nanojit-central/rev/49982fe529dd
Created attachment 450070 [details] [diff] [review] Quick fix to include signal.h. The previous patch used "raise" and "SIGABRT", but didn't include signal.h. This broke the nanojit-central build both on ARM/Ubuntu and x64/Ubuntu, so I've pushed a trivial patch to fix it. http://hg.mozilla.org/projects/nanojit-central/rev/0a8efeddb4d9