著者: Philip Hazel 日付: To: Christian Persch CC: pcre-dev 題目: Re: [pcre-dev] Security risk or not? Changing PCRE options from
patterns.
On Mon, 1 Oct 2012, Christian Persch wrote:
> Are any of the other ones dangerous? Afaict not. So limiting
> this new compile or runtime option's effect to (*UTF8) would be enough.
Of course, any application that is worried about this can itself check
for the text (*UTF8) at the start of any user pattern that it passes on
to PCRE. It could even use PCRE to do the check! To do it properly,
quite a complicated pattern is needed because other settings such as
(*CR) can precede (*UTF8) at the start of a pattern. Something like this
should be quite efficient:
^(?:\(\*\w+\))*?\(\*UTF\d+\)
It only needs to be used if the pattern begins with '(*', so for many
patterns the extra check will be insignificant.
There are only two bits left in the PCRE options definitions (out of
32), and I am rather reluctant to use one of them just for this check.
