Thanks Daniel. I'll check the CVS. I agree that "exponential" is a better word than "infinite loop", since during test I felt that the bigger the xsd file is, the (much) longer it takes. Sometimes when the file size is not too big, the problem will finally stop after running for a long time.
Best, and have a great holiday :-) Yong Chen > -----Original Message----- > From: Daniel Veillard [mailto:[EMAIL PROTECTED] > Sent: Wednesday, November 01, 2006 1:27 PM > To: Yong Chen (yongche) > Cc: [email protected] > Subject: Re: [xml] bug 362989 > > On Wed, Nov 01, 2006 at 12:12:11PM -0800, Yong Chen (yongche) wrote: > > > > Hi Daniel, > > Hi Yong, > > > Thanks a lot, I do appreciate your effort, time, and kindness. > > see the bitter frenchman sometimes actually helps. Today > was holliday in France, so I managed to get some time ! > > > Also wanted to let you know, it's not that I didn't want to > spend my > > time on it, you know I'm using libxml2 seriously. Actually > I've been > > debugging this issue for a long time, including reading > libxml source > > code, trimming down the xsd file for all kinds of testings. > Trust me, > > I did spend lots of time on it. > > Okay :-) What made it hard for me was that content for that > type had initially more than a hundred state, impossible to > track the behaviour completely, problem was to spot the > problem pattern. > > > As for my xsd file, I knew it had "non-deterministic" problem (that > > is, name uniqueness violation), I used a Java xsd validator and it > > showed me the problem. But libxml2 should show me the same error > > (instead, it went into infinite loop), that's something I > want to fix. > > okay, reasonnable. That bit of information could have been > useful though. > > > Yes the xsd file is auto-generated, your comments remind me > to make it > > more readable (or user-friendly). I'll see if I can do > something about > > it. > > > > Again, thanks a lot. I'll fix my xsd issue. And where can I see the > > bug fix and optimization code? (in the bugzilla)? > > Ah, right, the fix is in CVS. I don't think it's really a > loop issue, rather that a behaviour which was exponential > being reduced to something more acceptable. Though it seems > to have fixed 5 more tests in runsuite, maybe there was a > real error underneath. A complex issue of recreating a > transition using the normal API to avoid duplicate instead of > direct in-situ modification of the transition. > I need to update the bugzilla entry, I made the mail first > ... good point. > > Daniel > > -- > Red Hat Virtualization group http://redhat.com/virtualization/ > Daniel Veillard | virtualization library http://libvirt.org/ > [EMAIL PROTECTED] | libxml GNOME XML XSLT toolkit > http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search > engine http://rpmfind.net/ > _______________________________________________ xml mailing list, project page http://xmlsoft.org/ [email protected] http://mail.gnome.org/mailman/listinfo/xml
