PCRE doesn't like illegal Unicode sequences (it's explicitely
documented in pcreunicode(3) that they trigger undefined behaviour,
and the program may crash). Therefore, we always let PCRE check
the validity of both the pattern and the subject string.
However, when performing global matching, the subject string
can be checked only once: subsequent matches can safely skip the check
and avoid a huge performance hit of scanning the whole subject
string for each match (!).
This patch implements that behaviour internally -- it's still
not possible for the user to skip the sanity check. On large
subject strings, this gives a terrific performance benefit.
Change-Id: Ia44cf18782e07966c9cd6ec4ccfef081ed131763
Reviewed-by: Robin Burchell <robin+qt@viroteck.net>
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
5c9ba41f71