https://bugs.exim.org/show_bug.cgi?id=2030
--- Comment #3 from Philip Hazel <ph10@???> ---
No, you are right. I had thought about this some more myself, and came to the
same conclusion. I will add a new option (called PCRE2_ENDANCHORED unless I can
think of something better) that causes the match to fail (and therefore
backtrack) if the end of the pattern is reached before the end of the subject.
Using this with PCRE2_ANCHORED should give the result you want. I will add this
as an option that, like PCRE2_ANCHORED, can be used at compile time or match
time, but I suspect that (again like PCRE2_ANCHORED) using it at match time
will disable the use of JIT.
However, this will not be for some time. Once 10.23 is released (in a couple of
weeks' time) and has settled for a week or two, I will get on with integrating
a new version of pcre2_match() that I have been working on. This uses the heap
for backtracking instead of the stack, but in a more efficient way than the
existing --disable-stack-for-recursion option (in fact, it seems to run about
10% faster than the current non-heap code). That will take time, as there is a
lot of documentation and other associated stuff to update, but as 10.23 is
coming out in February, I won't be planning on another release till around
August at the earliest.
As well as getting rid of "stack exceeded" errors, which have been a source of
irritation since the start of PCRE, the new pcre2_match() also fixes at least
one obscure bug, and allows for backtracking into recursions such as (?2) which
makes it more Perl-compatible. And the code has had a useful clean-up.
--
You are receiving this mail because:
You are on the CC list for the bug.
