what's the motivation here?

Using insSkip for this was gross hack dmandelin and I did because we didn't have a better way to store the regexp source string. Now we do. This patch removes the last use of insSkip from TM. Tamarin doesn't use insSkip at all, so we can rip it out of NJ, reducing the complexity of LIR walking and saving another 100 lines of code or so.

This overlaps significantly with bug 495734.

Comment on attachment 397547 [details] [diff] [review] patch Looks good to me, however, dmandelin wrote this part, so I might not be the best one to review.

I forgot: perhaps you could use the assignment form of reference initialization to make + VMAllocator& alloc(*JS_TRACE_MONITOR(cx).reAllocator); easier for me to parse?

Running this on the patched code causes an assertion in: #1 0x0818338d in nanojit::Assembler::pageValidate (this=0x9ccdf54) at ../nanojit/Assembler.cpp:212 #2 0x0818763e in nanojit::Assembler::gen (this=0x9ccdf54, reader=0xbfa8a204) at ../nanojit/Assembler.cpp:1314 #3 0x081878eb in nanojit::Assembler::assemble (this=0x9ccdf54, frag=0x9cd0b64) at ../nanojit/Assembler.cpp:677 #4 0x08191a52 in nanojit::compile (assm=0x9ccdf54, frag=0x9cd0b64, alloc=@0xb7b08008, labels=0x9cd0674) at ../nanojit/LIR.cpp:2040 #5 0x0811ba32 in RegExpNativeCompiler::compile (this=0xbfa8a3c0, cx=0x9cd0e48) at ../jsregexp.cpp:3122

Obsolete with the removal of tracejit.