This may result in bizarre code generation behavior or crashes when compiling a UInt32ToNumber that must perform a double conversion.
Created attachment 109999 [details] the patch
Comment on attachment 109999 [details] the patch ouch
Can this be regression tested?
Landed in r96860.
(In reply to comment #3) > Can this be regression tested? I think we should strive to do this right, in one go: https://bugs.webkit.org/show_bug.cgi?id=69565 The DFG's speculation on arithmetic operations is starting to stabilize to the point where it can be succinctly described and elegantly tested. I think we should take a break at some point from adding new bugs to the DFG and figure out a testing harness that can do this the right way, and uniformly, to minimize the number of redundant tests that we add.