fixed JRuby schema validation when XSD has DOCTYPE to DTD

[patcheng]

consolidation of #791 and #810

[leejarvis]

This looks great, thanks for the quick response Patrick!

[pebrinic]

Awesome fix! Would it be possible to get a new RC built with this change?

[flavorjones]

1.5.10 will be released later today with this fix.

[ppg]

Hi,

I believe I'm still hitting an issue with this on Ubuntu 8.04 (and 10.04). Ruby-saml is trying to read the protocol XSD file mentioned in a related issue and its generating a segfault:

../shared/bundle/ruby/1.8/gems/nokogiri-1.5.10/lib/nokogiri/xml/schema.rb:37: [BUG] Segmentation fault
ruby 1.8.7 (2011-02-18 patchlevel 334) [i686-linux]

This is running through passenger 3.0.7 and fronted by apache2 (whatever the up-to-date package is on 8.04 and 10.04). Predictably it stops any saml authentication (i.e. through omniauth) from working.

The odd parts are:

• This works on my development machine (10.04) just fine, but not on an internal testing server (tried both 8.04 and 10.04; note this is a vmware virtual machine if that affects anything)
• If I run the Nokogiri::XML::Schema.new() call with the XSD file in IRB, it complains about the formatting of it (on both my development machine and the internal testing server which I also find wierd), but it doesn't generate a segfault.

So my questions are is this the same issue that wasn't fixed all the way, or am I seeing something different? And why the odd behavior where same OS, dependencies, etc. seem to behave differently on the two different machines? And how can I trace down further what is causing this issue and either fix it or workaround it?

Thanks,
\Peter

[ppg]

Hmm, so I didn't realize that the chdir from ruby-saml was important to be able to find the other schemas in the directory. Now that I have that working, my development machine correctly parses the schema and returns a Nokogiri::XML::Schema object in IRB. However, now the confusing part is that the internal test server also can parse it in IRB, but when running under passenger/apache its segfaulting? I feel better about it correctly parsing on my development machine, since it works correctly there, but I feel farther from a solution with less info because now I'm at a loss on how to figure out why its segfaulting? Any suggestions?

Thanks,
\Peter