|
Hi, > Damn, right, good spotting. :)
> You seems to have a really large test collection, are you using something > public or a home grown one ? It’s a mixture of both!
> That doesn't seems the right approach, by freeing cur, there you >change the semantic of the API, that doesn't look good to me. What should > be done instead is check the return value of xmlAddChild() and if NULL > then free the node in the caller. I think that call is done in SAX2.c > at the end of xmlSAX2Reference() . I made that change and the leak is indeed > fixed, I will commit this to SVN later.
I thought of fixing at the same place, but then decided to do it inside xmlAddChild because I thought it would take care of the problem at other places where the return value is not checked, but I wasn’t really sure whether it was the right way to go about it.
> Ideally all xmlAddChild() call in the library should be checked in that > way, to avoid such problems. Maybe you could look at this at least in the > most critical parts like the SAX2 and Include modules. That's my next target ;)
And a big thanks to you and libxml, for the continued co-operation and help.
Regards Ashwin |