re_intuit_start(): MBOL use char for float max

Do the "maximum place \n can appear within the float range" calculation
in chars rather than bytes. Doing it in bytes is logically incorrect,
although I think the worst outcome is that a string is falsely accepted
by intuit then has to be failed by a full run of the regex engine.
But I couldn't think of a test that would show a significant performance
difference.

diff --git a/regexec.c b/regexec.c
index f9cfc89..213bf9f 100644 (file)
--- a/regexec.c
+++ b/regexec.c
@@ -1112,9 +1112,12 @@ Perl_re_intuit_start(pTHX_
             goto fail_finish;
         }
 
+        /* since *t == '\n', it's safe to do +1 here rather than HOP(t,1) */
         rx_origin = t + 1; /* earliest possible origin is after the \n */
 
-        if (t >= check_at - prog->check_offset_min) {
+        if (prog->substrs->check_ix == 0  /* check is anchored */
+            || t >= HOP3c(check_at,  - prog->check_offset_min, strpos))
+        {
             /* Position contradicts check-string; either because
              * check was anchored (and thus has no wiggle room),
              * or check was float and t is above the float range */