On Wed, 22 Feb 2012, Giuseppe D'Angelo wrote: > I just saw the change in the commit log: > > Set PCRE_EXTRA_USED_JIT when JIT was actually used at runtime > > So here's the stupid question of the day: doesn't that break > thread-safety? Now using the same pcre_extra block in multiple threads > causes races on who sets/clears that flag.
Not a stupid question. Thanks for asking it. Darn. I want this functionality for pcretest. I do not want to modify the API for pcre_exec. Is there any other way I can achieve this? (I was slightly worried when I implemented what I did, but obviously not worried enough. I must remember to *think threads*. :-) OK, here is an alternative idea: if the offsets vector is large enough, put the information in the slot after the last slot that is used. This *could* be made conditional on an option, but (a) it does no harm, and (b) there are only two option bits left. It's a bit kludgy, but it does the job. (The PCRE API is already rather creaky - in the very long term it may be necessary to re-design it completely, in which case this kind of kludge could be tidied up.) What do you think? Philip -- Philip Hazel -- ## List details at https://lists.exim.org/mailman/listinfo/pcre-dev
