On 07/06/2016 09:14 PM, arn...@skeeve.com wrote:
Can you explain the story behind this? Why is support for non-ASCII ranges being pulled out of dfa? I am concerned that this will affect gawk.
This new change shouldn't affect gawk, since that horse left the barn a couple of years ago (grep commit 8b16e90a248d70aef687931370f8ef7c1f2bc4ef dated 2014-02-27). Because of that earlier change, when dfa.c sees a non-ASCII range it inserts a pattern that will bail out (just as it would for a back-reference), and the caller is then supposed to fall back on a backtracking matcher. I assume Gawk does this already.
The new change shouldn't alter the externally-visible behavior of dfa.c's entry points; it should merely simplify the dfa.c code given that dfa.c no longer has to worry about non-ASCII ranges etc.
