Re: [xml] bug 362989

[Daniel Veillard <veillard redhat com>]

On Tue, Oct 31, 2006 at 08:00:59PM -0500, Daniel Veillard wrote:
> On Tue, Oct 31, 2006 at 02:25:49PM -0800, Yong Chen (yongche) wrote:
> This is the epsilon transition elimination in the graph of transition.
> Each time someone tried to point out that the implementation of this function
> was broken it was something else, it's just where the problem usually show
> up, the error being the construction of the graph automata. Of course the
> construct in the automata are a bit more complex than in the book.
>   It is not trivial, it takes serious concentration and time to really
> understand where the problem comes from. The bigger the example, the bigger
> the automatas, the more time it takes. You may start to understand why I insist
> on getting a reduced problem.

  I spent 3 hours on the problem, a large part of it trying to understand
your schemas. And I found a bug, really not obvious, and a couple of
optimizations.

>   I would also like to raise the point of the 'niceness' of that schemas.
> It seems to me trying to understand the semantic of that is near impossible,
>   <xs:group ref="optional_group_4" minOccurs="0" maxOccurs="1"/>
> looks like automatically generated schemas with names without semantic
> constant references to labels, making even harder to get the full picture.
> It's a bit like asking to debug code which was automatically generated
> i.e. even more painful than when you can extract the semantic of the
> code.

  I guess you will appreciate the fact that I think I fixed the bug:

paphio:~/XML -> xmllint --timing --noout --schema 362989.xsd test.xml
362989.xsd:23: element complexType: Schemas parser error : complex type 'access-list_type': The content model is not determinist.
WXS schema 362989.xsd failed to compile
Compiling the schemas took 1195 ms
Parsing took 0 ms
Freeing took 0 ms
paphio:~/XML -> 

  I guess you now need to fix your schemas, it it not deterministic, and
hence should be rejected by any correct XSD implementation. Let's say I
really did *not* enjoy debugging in your schemas, still I'm kind and will
give you the reason leading to my believing it's broken:
    - see the content model for access-list_type
    - notice how all groups direct or derived use minOccurs="0"
    - notice how all elements in the content model also have minOccurs="0"
    - notice that element "src_addr" is defined by both
      block_group_18 and block_group_3
    - block_group_18 is referenced by block_group_17
    - block_group_3 is referenced by block_group_2
    - block_group_2 is referenced by access-list_type
    - block_group_17 is referenced by access-list_type

 so this can't be determinist, your schemas is broken

Daniel

-- 
Red Hat Virtualization group http://redhat.com/virtualization/
Daniel Veillard      | virtualization library  http://libvirt.org/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine  http://rpmfind.net/