Re: [xml] XPath on a subtree

[Daniel Veillard <veillard redhat com>]

On Sun, Mar 31, 2013 at 06:44:44PM +0800, Daniel Veillard wrote:
> On Sat, Mar 30, 2013 at 07:11:35PM +0000, Alex Bligh wrote:
> > > BTW the xml is generated so that should be discarded from the
> > > patch,
> >
> > That's strange. It appears in git and has a history.
> >   https://git.gnome.org/browse/libxml2/tree/doc/libxml2-api.xml
> >
> > How does one regenerate it (as the normal build process
> > didn't)?
>
>   Ah, well usually I do this only at release times, but it can
> be useful when developping a patch too, so have a look at
> doc/apibuild.py
>
>   It is a large bit of rather ugly Python, but it scans the .h
> files, the .c files, and generate a formal API description of
> all exported entry points and structures:
>   doc/libxml2-api.xml
>
> > From there a number of other things are generated such as the
> online docs, checks for the exported symbol files, python
> bindings ...

  Ah, another thing which is generated from it is testapi.c
(make rebuild_testapi) which is used to exercise the full set of
the API against weird input etc ...

....
Testing xpath : 32 of 40 functions ...

Program received signal SIGSEGV, Segmentation fault.
0x00000000004f9691 in xmlXPathSetContextNode (node=0x82bd50, ctx=0x0)
    at xpath.c:15091
15091       if (node->doc == ctx->doc)

 and it indicates you forgot to check the arguments from that new
function. There is actually a number of possible error case so
it is better to make the function return an error status,
I also modified xmlXPathNodeEval to return NULL if str is NULL or
if xmlXPathSetContextNode failed :-)

I commited the result:

  https://git.gnome.org/browse/libxml2/commit/?id=28876afb4e67f9f4292fc2a7134f6b97f0618b7d

 thanks !

Daniel

-- 
Daniel Veillard      | Open Source and Standards, Red Hat
veillard redhat com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | virtualization library  http://libvirt.org/