Make it works!
Created attachment 99610 [details] patch
Comment on attachment 99610 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=99610&action=review > Source/JavaScriptCore/ChangeLog:8 > + Move the generated code to the .text section instead of .data section. Nice catch! Would it make sense to try this optimization on ARM?
> Would it make sense to try this optimization on ARM? We need to detect whether hard or softfp ABI is present. Hopefully the compiler will tell us...
Comment on attachment 99610 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=99610&action=review > Source/JavaScriptCore/jit/ThunkGenerators.cpp:142 > + "jmp " SYMBOL_STRING_RELOCATION(function) "\n" \ Changing this to a jump will misalign the stack on x86_32/mac are you sure it's not complaining about it?
> Changing this to a jump will misalign the stack on x86_32/mac are you sure it's not complaining about it? Misalign? Why? Keeps 8 byte alignment, even if you don't push an extra 8 byte return address to it. Or is that 16 byte aligned? I don't know about mac ABI.
(In reply to comment #5) > > Changing this to a jump will misalign the stack on x86_32/mac are you sure it's not complaining about it? > > Misalign? Why? Keeps 8 byte alignment, even if you don't push an extra 8 byte return address to it. Or is that 16 byte aligned? I don't know about mac ABI. Mac ABI requires 16byte alignment
> Mac ABI requires 16byte alignment thanks for letting me know. Tomorrow I will put it back and resubmit the patch.
Created attachment 99685 [details] updated patch
Seems reasonable to me, but the real JSC experts shoudl comment.
Comment on attachment 99685 [details] updated patch Clearing flags on attachment: 99685 Committed r90425: <http://trac.webkit.org/changeset/90425>
All reviewed patches have been landed. Closing bug.