W3C home > Mailing lists > Public > public-qa-dev@w3.org > June 2010

Re: Validator timeout and XML-LibXML bug

From: Ville Skyttä <ville.skytta@iki.fi>
Date: Fri, 11 Jun 2010 01:34:24 +0300
To: "Dominique Hazael-Massieux" <dom@w3.org>
Cc: public-qa-dev@w3.org
Message-Id: <201006110134.25064.ville.skytta@iki.fi>
On Thursday 10 June 2010, Ville Skyttä wrote:

> One problem with the XML::LibXML::InputCallback way documented next to
> ext_ent_handler is that *all* URIs, including the URI to the actual
> document to be parsed get passed to it, and that URI won't ever be within
> our sgml-lib jail.  But it just occurred to me that this might not
> actually be a problem as we pass the doc to be validated as a string
> containing the document and not a filename/URI.  I'll experiment more and
> report back.

Candidate fix is now in CVS, seems to fix the problem on my box and qa-dev:
http://dev.w3.org/cvsweb/validator/httpd/cgi-bin/check.diff?r1=1.776&r2=1.777
(Needs also catalog.xml which is also in CVS).

I don't know how easy it would be port just this change over 0.8.6.  Regarding 
status of code in CVS HEAD, this patch I committed earlier after 0.8.6 needs 
to be rechecked:
http://dev.w3.org/cvsweb/validator/httpd/cgi-bin/check.diff?r1=1.768&r2=1.769
It may have been premature to drop support for non-structured errors, the 
XML::LibXML docs still seem to indicate that non-structured ones may still be 
thrown in some cases (no idea in exactly which) with 1.70, and if that's the 
case, the above should be pretty much reverted.

Sigh.  The more I work with XML::LibXML, the less I like it :(
Received on Thursday, 10 June 2010 22:35:04 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 19 August 2010 18:12:51 GMT