On Mon, 08 Aug 2016 17:34:57 -0700, timo wrote: > to be more precise, the way we code-gen "literal" qregex nodes with > subtype "ignoremark+ignorecase" will only ever check the ordbaseat of > the first character in the literal against the haystack. >
This has been fixed as of https://github.com/perl6/nqp/commit/18e40936a on nqp, and the regex nodes cleaned up. With the new nqp::indexicim and eqaticim ops in MoarVM we should have greater maintainability of that code and less difference between the different types of regex, mostly using the same code now. The rakudo version where I bumped nqp is 2017.05-381-g1ac7996a4 Tests: I have added tests to roast here, https://github.com/perl6/roast/commit/0a80d0d2e84c824d0d5cfde1bbf40440ba0732d8