Re: [pcre-dev] External memory allocator interface

On Tue, 10 Jun 2008, Roman Rybalko wrote:

> Here it is. (just a draft)

That was quick. What I didn't tell you before was that I have just been
away for 2.5 weeks, and I am going away again for 10 days at the end of
this week. Because of this, I do not have any time this week to look at
your code in any detail until some time in July at the earliest.

> I've ported pcre_compile* and pcre_exec, but NO_RECURSE is broken now,
> at least because I don't know yet how to deal with it in the new
> context. Probably I should add stack_malloc/stack_free methods to struct
> pcre_mm.

I took a *very* quick look. One thing that struck me was that there
would be fewer changes, and the performance would be less impacted, if
instead of adding the mm argument to the match() function, you added an
extra field to the match_data structure. That is generally where
"static" data for the functions is placed.

> Generally, I see all this as a new MM layer on top of existing.

That seems like a good approach.

Philip

--
Philip Hazel