I was getting disturbingly frequent but intermittent failures with rooting analysis, and thought I'd dig in a little.

I know we're going to be ripping out the dynamic rooting analysis soon, but this gets rid of a lot of intermittent orange for me. But certainly not all -- I have a lingering very-intermittent: Assertion failure: masm.framePushed() == initialStack, at /home/sfink/src/MI-GC/js/src/jit/IonCaches.cpp:975 even with this patch applied.

Comment on attachment 816972 [details] [diff] [review] Prevent executable allocator from handing out poisoned pointers with JSGC_ROOT_ANALYSIS Review of attachment 816972 [details] [diff] [review]: ----------------------------------------------------------------- r=me ::: js/src/assembler/jit/ExecutableAllocatorPosix.cpp @@ +46,5 @@ > + void* allocation; > +#ifdef JSGC_ROOT_ANALYSIS > + do { > +#endif > + allocation = mmap(NULL, n, INITIAL_PROTECTION_FLAGS, MAP_PRIVATE | MAP_ANON, VM_TAG_FOR_EXECUTABLEALLOCATOR_MEMORY, 0); The indentation here looks wrong to me. I guess the style guide doesn't really cover this case, but I would suggest indenting this line. It doesn't really matter though.