Re: [xml] Improper Type Returned For Repeated Attribute Declaration

[Daniel Veillard <veillard redhat com>]

On Thu, Jan 10, 2008 at 01:49:42AM +0530, ashwin sinha wrote:
>    On  Jan  8,  2008  8:26  PM,  Daniel Veillard <[1]veillard redhat com>
>    wrote:
>
>    On Tue, Jan 08, 2008 at 08:02:29PM +0530, Ashwin wrote:
>    >
>    >    Hi,
>    >
>    >         In  the  attached  file  the  attribute a1 is declared twice
>    in the
>    >      internal   subset.  The  draft  (XML 1.0) says that this is ok,
>    however it
>    >      is binding that the type of the attribute value should be taken
>    as the
>    >      first  one  which  occurs in the definition, in this case CDATA
>    occurs
>    >      before   Nmtokens   therefore  the  type of a1 should be CDATA.
>    However,
>    >     this  is  not  the  case, while parsing the type associated with
>    a1 is
>    >      NMTOKENS  and  normalization  is  performed  on  the  attribute
>    value
>    >      according  to  the  rules  specified in the XML draft. However,
>    ideally in
>    >      this  case  there  should  be no normalization and a1 should be
>    treated as
>    >    CDATA.
>
>        >Hum, yes there is a problem in this edge case:
>         >  -  You will notice that when saving the document xmllint will
>      not
>         >   output  the  superfluous NMTOKENS definition in the internal
>      subset,
>        >   so it seems libxml2 does not store it, which looks good
>        > - but as you noted the attribute value is normalized and that's
>      not
>         >    okay.  If  one  comment  out the NMTOKENS definition in the
>      internal subset
>         >   then  the  value  is  not  normalized, which is the expected
>      behaviour.
>         >So  for  some reason the NMTOKENS information is not completely
>      dropped
>      >  at  parse  time  and  retained  and influence the parsing of the
>      attribute.
>      >  I  guess using a debugger and following the attribute processing
>      in the
>      >  start tag SAX handler should allow to find this out quickly, I'm
>      on
>      >  the road at the moment, not sure I will be able to spend time on
>      it quickly
>      > but if you can look for this it would be nice to get this fixed,
>      >  thanks  for the report, ideally the document should be stored in
>      the regression
>      > suite once the bug is fixed,
>
>
>       Hi,
>          The  above  problem occurs because on encountering the second a1
>    attribute, even though its a duplicate the parser still adds it to the
>    hash list using xmlAddSpecialAttr, and then while retrieving using the
>    hash  key  it  fetches  the  most  recent  entry  of  a1 from the hash
>    list(which happens to be the 2nd Nmtoken a1 entry.)
>
>          I  am  hoping to fix the problem by adding a global flag which i
>    will  set  in case a duplicate attr value is enountered (the check for
>    duplicate  attr  value is being done in valid.c), and adding this flag
>    to  the  if  check  before  calling the xmlAddSpecialAttr function. Is
>    there any potiential problem with this fix?

  yes global state is really a problem in a library, this should really
be avoided, I don't think this scheme can work well, instead the error
of duplication should be caught, andpropagated back to the place where
xmlAddSpecialAttr is called.
  I will look at this probably tomorrow morning,

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/