On Fri, 09 Jun 2006 21:07:47 +0100, Dan Janowski <[EMAIL PROTECTED]> wrote:
> Hi,
>
> Has there been a discussion about the memory-model changes? If so, I
> am interested to understand the direction better. Is there a
> particular list thread that I might read?
>
Sean and I have discussed it a little bit, though it must have been
offlist since I can't find it in the archive :I . There are a few threads
about problems / bugs caused by memory handling, e.g. the segfault bug
Doug Bryant reported recently:
http://rubyforge.org/pipermail/libxml-devel/2006-June/000162.html
> In my past ruby extensions I have avoided giving the GC any more than
> necessary, to limit memory copying as well as performance. Is the
> libXML memory model unstable to keep pointer references to?
>
Not especially, there are just a few special cases that we don't currently
cater for, such as when copying nodes between documents. If a node pointer
is added to a second document, and them xmlFreeDoc is called on the first,
that pointer becomes invalid and will cause a segfault either when the
document is accessed, or when xmlFreeDoc is called on it. So in this case,
I think we do need to do a recursive copy of the nodes, so that they can
be garbage collected with the new document. I'm still learning the
intricacies of libxml2 but this is how it looks to me right now.
I'm not planning to copy indiscrimiately, however - pointers will probably
be retained to nodes copied within the same document, and with attributes
wherever possible. Mainly I'm looking at rationalizing our memory
handling, cleaning up and making sure we're using the right
ruby_xml_whatever_new[123 etc.] function in the right place.
--
Ross Bamford - [EMAIL PROTECTED]
_______________________________________________
libxml-devel mailing list
libxml-devel@rubyforge.org
http://rubyforge.org/mailman/listinfo/libxml-devel
