Nokogiri::XML::SAX::PushParser's native_write in JRuby...

[kitplummer]

doesn't seem to like the header, while in regular Ruby it doesn't have a problem. I'm guessing this is a Xerces vs. libxml issue, but wondered if anybody else has seen this issue.

I'm using 1.5.0.

I've captured a simple script that shows the error returned by the fog.io library:

https://gist.github.com/1328982

Not sure if this is a bug, or something that should be handled in the next abstraction up. Thoughts/ideas?

[yokolet]

Hello.

I got authentication related errors on both Ruby 1.9.2 and JRuby 1.6.5 when I ran the code above. So, I couldn't reproduce the same error as yours.

Can you provide the XML document that fog tried to parse?

[kitplummer]

Ah, I had removed the appropriate key_id and key, sorry for that.

<?xml version="1.0" encoding="UTF-8"?> <RunInstancesResponse xmlns="http://ec2.amazonaws.com/doc/2011-05-15/">     <requestId>b9aea9d2-0d95-4129-94cb-2c2f9f22e8fb</requestId>     <reservationId>r-6a522804</reservationId>     <ownerId>077892567685</ownerId>     <groupSet>         <item>             <groupId>sg-ec066985</groupId>             <groupName>quick-start-1</groupName>         </item>     </groupSet>     <instancesSet>         <item>             <instanceId>i-7318c610</instanceId>             <imageId>ami-195e9170</imageId>             <instanceState>                 <code>0</code>                 <name>pending</name>             </instanceState>             <privateDnsName/>             <dnsName/>             <reason/>             <keyName>lucee-demo-keypair</keyName>             <amiLaunchIndex>0</amiLaunchIndex>             <productCodes/>             <instanceType>t1.micro</instanceType>             <launchTime>2011-11-01T22:46:22.000Z</launchTime>             <placement>                 <availabilityZone>us-east-1a</availabilityZone>                 <groupName/>                 <tenancy>default</tenancy>             </placement>             <kernelId>aki-08ed0761</kernelId>             <monitoring>                 <state>disabled</state>             </monitoring>             <groupSet>                 <item>                     <groupId>sg-ec066985</groupId>                     <groupName>quick-start-1</groupName>                 </item>             </groupSet>             <stateReason>                 <code>pending</code>                 <message>pending</message>             </stateReason>             <rootDeviceType>ebs</rootDeviceType>             <rootDeviceName>/dev/sda1</rootDeviceName>             <blockDeviceMapping/>             <virtualizationType>paravirtual</virtualizationType>             <clientToken/>             <hypervisor>xen</hypervisor>         </item>     </instancesSet> </RunInstancesResponse>

[yokolet]

Thank you. Now, I got the exactly the same error. However, libxml version also resported the Syntax error. Perhaps, copy & paste didn't do correctly?

Anyway, I confirmed pure Java version has a bug of push parser. I used some of Nokogiri's test xml files to see how push parser (without fog) worked. "po.xml" raised "Nokogiri::XML::SyntaxError: The processing instruction target matching "[xX][mM][lL]" is not allowed." on pure Java version. But, not on libxml version. I'll look into this.

For reference, the test code is https://gist.github.com/1333808
I ran this code from Nokogiri's top directory as in below:

jruby -Ilib -Itest ../Canna/src/snippet/issue559_test.rb tesfiles/po.xml

[kitplummer]

Any news on this? Figured I'd ask before I start debugging myself. :)

[yokolet]

Sorry that I took long time to get back to you.

I rewrote push parser. Now, I got the result below from the gist, https://gist.github.com/1333808

jruby -Ilib -Itest ../Canna/src/snippet/issue559_test.rb test/files/po.xml 
/Users/yoko/Documents/workspace/Ranunculus/test/helper.rb:10: version info: {"warnings"=>[], "nokogiri"=>"1.5.3.rc3", "ruby"=>{"version"=>"1.9.2", "platform"=>"java", "description"=>"jruby 1.6.8.dev (ruby-1.9.2-p312) (2012-04-11 03c14ba) (Java HotSpot(TM) 64-Bit Server VM 1.6.0_29) [darwin-x86_64-java]", "engine"=>"jruby", "jruby"=>"1.6.8.dev"}}
[["purchaseOrder", [["orderDate", "1999-10-20"]]], ["shipTo", [["country", "US"]]], ["name", []], ["street", []], ["city", []], ["state", []], ["zip", []], ["billTo", [["country", "US"]]], ["name", []], ["street", []]]
Loaded suite ../Canna/src/snippet/issue559_test
Started

Finished in 0.015000 seconds.

0 tests, 0 assertions, 0 failures, 0 errors, 0 skips

Test run options: --seed 57946

Does this mean push parser works for you?

Also, please look at my comment in the similar issue, https://github.com/tenderlove/nokogiri/issues/615 .

If you have a chance, would you try master?

[kitplummer]

I'm still seeing the same original error, when using 1.5.3.rc3.

[yokolet]

Can you check your XML document doesn't have anything before XML declaration? If there's a space, "\n" , or something before decl, Xerces complains that.

[kitplummer]

Of course. I'll dive into this tonight...hopefully it's just something stupid on my side. Thanks for the effort.

[yokolet]

Hmmm, OK. There should be something wrong in pure Java Nokogiri. I'll try your gist with fog.

[jvshahid]

I can't reproduce this bug anymore. If you can still repro it, please reopen the issue.