Re: [xml] FW: Some more memleak fixes in Parser.c

[Daniel Veillard <veillard redhat com>]

On Thu, Apr 24, 2008 at 05:12:48PM +0800, Ashwin wrote:
> > While the description sounds good, the patch didn't apply to CVS, could 
> > you double-check the patch please ?
>       
> Hi,
>    Sorry for the delay in reply. I didn't understand, you mean the patch was
> wrong? Anyway I am attaching patch again, please let me know if it is ok...
> Basically checking if a mem error has occurred in inputPush and returning
> from there if it has....
>
> > Another query is for a lot of places where Inputpush function is invoked
> the
> > return value is not checked, from what I could make out this is because at
> > the time of context creation the number of input streams is set to 5, and
> > since push is being called directly afterwards, the number of streams will
> > not be more than 5 and thus the realloc block inside inputPush will not be
> > hit. So perhaps that is why memory failure is not expected in those paths.
>
>
> > >  Hum, no, I guess it's an oversight which should be fixed too.
>
>    Inside inputPush function if realloc fails I am now freeing the
> inputstream, however I am not 100% sure this is the correct way to go about
> it... 

  Another old patch pending, yes this need to be fixed, I just changed
a couple of things:
   - made xmlPushinput return an int, that's a compatible change
   - made pushInput return -1 on error instead of -1 like all other push
     functions, none of libxml2 code was testing the return value anyway
     because this was broken, 0 is a legal value which you should get
     when pushing the first entity.

Some further cleanup would be to check inputPush() return value
everywhere, and that's around 30 places within libxml2. Most of them
though are used for the main entity, so I guess if it fails the failure
is caught pretty soon as there is nothing to parse !

  thanks for the patch, sorry for the delay,

Daniel

-- 
Daniel Veillard      | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
daniel veillard com  | Rpmfind RPM search engine http://rpmfind.net/
http://veillard.com/ | virtualization library  http://libvirt.org/