Re: [APC-DEV] Recursive zval patch for review/comments

Rasmus Lerdorf wrote:

>> The real problem is that sma_deallocate() acts rather dumb about
>> whether it got a valid offset to free or not.
> 
> Good find Gopal.  

Checking in a fair bit of canary checks into the non-debug codebase -
it will abort the process if some junk data (like an emalloc'd ptr)
is passed to deallocate. If in case that does happen somewhere, should
give me a nice backtrace in gdb.

Does make the code a fair bit slower, but overall APC is much faster
today than it was last week - and hopefully won't segv as much.

> Can we try to aim for a public release in the next couple of weeks?

Sure. If we leave spinlocks and optimizer out of this release, I
suppose.

Cheers,
Gopal
-- 
The difference between insanity and genius is measured by success