Re: [xml] DTD validation not restrictive enough?

[Morus Walter <morus walter tanto-xipolis de>]

I hope it's ok, if I put this on the list again - you replied to me
directly.

Jarek Dukat writes:
> Morus Walter wrote:
>
> > it's most probably due to an inapropriate usage of the library.
>
> My code is very simple:
>
> xmlDoc *doc = xmlParseFile("foo.xml");
> if (doc == NULL) {
>    // ...
> } else {
>    mlDtdPtr dtd = xmlParseDTD(0, (unsigned char*)"cfg.dtd");
>    xmlValidCtxt cvp;
>    cvp.userData = (void *) stderr;
>    cvp.error = (xmlValidityErrorFunc) fprintf;
>    cvp.warning = (xmlValidityWarningFunc) fprintf;
>    if ( !xmlValidateDtd(&cvp, doc, dtd) ) {
>      // ...
>    }
>    // ...
> }
>
> > I'm not familiar enough with the C API to guess what might be wrong in your
> > code, but you might have a look at xmllints sources, to see how validation 
> > is done.
>
> xmllint sources are very complex, buf if noone suggests simple solution 
> I'll have to take a closer look at it. Thanks.
the related section does not look very different to what you do,
except the initialisation of the xmlValidCtxt:
    /*
     * A posteriori validation test
     */
    if ((dtdvalid != NULL) || (dtdvalidfpi != NULL)) {
        xmlDtdPtr dtd;

        if ((timing) && (!repeat)) {
            startTimer();
        }
        if (dtdvalid != NULL)
            dtd = xmlParseDTD(NULL, (const xmlChar *)dtdvalid); 
        else
            dtd = xmlParseDTD((const xmlChar *)dtdvalidfpi, NULL); 
        if ((timing) && (!repeat)) {
            endTimer("Parsing DTD");
        }
        if (dtd == NULL) {
            if (dtdvalid != NULL)
                xmlGenericError(xmlGenericErrorContext,
                        "Could not parse DTD %s\n", dtdvalid);
            else
                xmlGenericError(xmlGenericErrorContext,
                        "Could not parse DTD %s\n", dtdvalidfpi);
            progresult = 2;
        } else {
            xmlValidCtxtPtr cvp;

            if ((cvp = xmlNewValidCtxt()) == NULL) {
                xmlGenericError(xmlGenericErrorContext,
                        "Couldn't allocate validation context\n");
                exit(-1);
            }
            cvp->userData = (void *) stderr;
            cvp->error    = (xmlValidityErrorFunc) fprintf;
            cvp->warning  = (xmlValidityWarningFunc) fprintf;

            if ((timing) && (!repeat)) {
                startTimer();
            }
            if (!xmlValidateDtd(cvp, doc, dtd)) {
                if (dtdvalid != NULL)
                    xmlGenericError(xmlGenericErrorContext,
                            "Document %s does not validate against %s\n",
                            filename, dtdvalid);
                else
                    xmlGenericError(xmlGenericErrorContext,
                            "Document %s does not validate against %s\n",
                            filename, dtdvalidfpi);
                progresult = 3;
            }
            if ((timing) && (!repeat)) {
                endTimer("Validating against DTD");
            }
            xmlFreeValidCtxt(cvp);
            xmlFreeDtd(dtd);
        }
    }

xmlNewValidCtxt just initializes the structure with 0's, something you're
missing, when you allocate the structure on the stack (at least if I remember
C programming correctly).

Morus