https://bugs.exim.org/show_bug.cgi?id=2555
--- Comment #5 from Giuseppe D'Angelo <dangelog@???> ---
(In reply to Philip Hazel from comment #4)
> The input to pcre2test is always 8-bit. If the utf option is set, that
> pattern and its test data are interpreted as UTF-8 characters. If you set
> the -16 or -32 options, the characters that are defined in UTF-8 are
> converted to UTF-16 or UTF-32 before being passed to the relevant PCRE2
> library. Thus, the same input works for all 3 libraries.
OK, thanks for confirming. Then indeed the original attachment is OK (it's the
UTF-16 pattern I had from fuzzing, converted and saved in UTF-8).
> This particular pattern fails fast in 8-bit mode because it is too big. If
> you compile --with-link-size=3 the 8-bit pattern behaves in the same way as
> the 16- and 32-bit patterns - that is, it is accepted but JIT takes a long
> time.
Great catch! I didn't think of trying a different link size.
So, given the above: is it deemed a problem or not that merely (JIT)compiling a
pattern can take a long time? Is the user expected to never compile/validate an
untrusted regexp?
--
You are receiving this mail because:
You are on the CC list for the bug.
