Re: [xml] Memory leak / cleanup missing after node creation.

[Jose Commins <axora myrealbox com>]

	Thanks again Igor!  During the process I also found some more memory 
oddness - the following below doesn't register as a leak using 'leaks' 
in FreeBSD Terminal, but I see the memory go up:

        In my (multithreaded) app I have x amount of buffers allocated with:

        theFinalXML = xmlBufferCreateSize(8192);

	If, for example, I perform a 'xmlBufferLength(theFinalXML)' I get back 
zero.  Er, OK, weird, since I thought 'xmlBufferCreateSize' can 
pre-allocate a known size so the buffer doesn't need to be constantly 
re-malloc'ed (or whatever scheme it uses).

        So if I dump the buffer (this is in my main thread loop):

        theFinalXMLSize = xmlNodeDump(theFinalXML, doc, root_element, 0, 0);

        I *have* to perform:

        xmlBufferEmpty(theFinalXML);

	afterwards, otherwise the buffers grow and grow.  Um, I am fine with 
using BufferEmpty (apart from the performace hit of constant allocs and 
reallocs), but then what use has 'xmlBufferCreateSize' ?


Regards,
                Jose.

On 24 May 2004, at 5:30 pm, Igor Zlatkovic wrote:

> On 24/05/04 17:59, Jose Commins wrote:
> >     By the way,
> >                     newNode = xmlNewText(theTagReplacements[1]);
> >                     xmlReplaceNode(cur_node, newNode);
> >     It still looks like a list of 'newNode's still has to be created 
> > and freed with 'xmlFreeNode' after 'xmlFreeDoc' is called; despite 
> > being assigned to the doc, xmlFreeDoc doesn't seem to remove the 
> > allocations for those 'newNode' s.
>
> That's probably because a new text node should be created in the 
> context of the document. Try
>
>   newNode = xmlNewDocText(cur_node->doc, theTagReplacements[1]);
>
> That should pepper it to fit most tastes.
>
> Ciao,
> Igor