We have some code to perform out of bounds checks in non-ASan builds, ASan checks would complement that. Built-in checks are better, because they detect all overruns and underruns, while ASan only detects small ones, which make Vector access a memory address in a redzone. Also, they can be made either debug only or release too, depending on performance requirements. ASan checks are better, because they check even those memory accesses that are performed via a raw pointer returned from data() or begin(). They have zero performance cost in non-ASan builds.
rdar://problem/19437718
Created attachment 244913 [details] proposed patch
Attachment 244913 [details] did not pass style-queue: ERROR: Source/WTF/wtf/Vector.h:510: Tests for true/false, null/non-null, and zero/non-zero should all be done without equality comparisons. [readability/comparison_to_zero] [5] Total errors found: 1 in 3 files If any of these errors are false positives, please file a bug against check-webkit-style.
Comment on attachment 244913 [details] proposed patch View in context: https://bugs.webkit.org/attachment.cgi?id=244913&action=review Do you know why the Win EWS bot is red on this patch? > Source/WTF/wtf/Vector.h:756 > + if (this == &other) // ASan will crash if we try to restrict access to the same buffer twice. > + return; It’s irritating to add this extra branch for all builds just to make ASAN work, especially since it only optimizes something that we should never be doing. Can we do better? Maybe just put this inside #if ASAN_ENABLED?
I don't yet understand why Windows failed to build; will need to figure it out before landing. > Maybe just put this inside #if ASAN_ENABLED? Will do.
Committed <http://trac.webkit.org/r178722>. Windows build was failing because of an incorrect macro value. Changed ASAN_ENABLED fallback value from false to 0 to fix this.