[RFC][icedtea-web] Fix PR955: regression: SweetHome3D fails to run

Danesh Dadachanji ddadacha at redhat.com
Wed Aug 1 08:13:44 PDT 2012


Here is a fix for PR955. [1]

This was quite a tricky patch to figure out. Proprietary webstart handles locales.. interestingly.. I've made ours use a fallback 
mechanism to find any information in the information tag. For each item it searches for (e.g. the title), it will try to find an 
information tag which has the closest possible locale that looks like your JVM's locale. Please see the unit tests for examples.

I have changed a few other things too with how we setup locales, the variant was not working properly. It was assumed to be 2 
characters long which is definitely not the case. My locale is en_CA.utf8 <-- variant has 4 chars.

My changes to the localeMatches method cover both the implementation I needed (being able to match multiple a locale by a given 
precision - the enum), as well as making it more easily understood. I hope the switch cases do this, it was a bit of a brain teaser 
figuring out those if statements and double negatives. :/

All other changes are in the ChangeLog below:
+2012-08-01  Danesh Dadachanji  <ddadacha at redhat.com>
+       Fix PR955: regression: SweetHome3D fails to run
+       * NEWS: Added entry for PR955
+       * netx/net/sourceforge/jnlp/JNLPFile.java: New enum Match that represents
+       the level of precision to use when matching locales.
+       (localMatches): Renamed to localeMatches, added matchLevel paramater
+       and updated conditionals to handle the level of precision specified by it.
+       (getVendor): New method that returns an information's vendor text.
+       (getInformation): Added override methods for getTitle and getVendor
+       that are used by the anonymous class to filter by locale. All three
+       methods now go through all levels of precision to search for the best
+       fitted locale.
+       (getResources), (getResourcesDescs): Updated to check if any level of
+       precision matches when searching for locales.
+       (parse): Added call to checkForTitleVendor.
+       * netx/net/sourceforge/jnlp/Parser.java
+       (checkForTitleVendor): New method to check for availability of localized
+       title and vendor from the information tags. Throws ParseException.
+       (getInfo): Replace loop with foreach loop.
+       (getInformationDesc): Remove check for present title and vendor.
+       (getLocale): Variant returned can now use everything after the eigth
+       element of the locale's string.
+       * netx/net/sourceforge/jnlp/resources/Messages.properties:
+       Update missing title and vendor messages to mention localization.
+       * tests/reproducers/simple/InformationTitleVendorParser/testcases/InformationTitleVendorParserTest.java:
+       Update output string as per new changes to Messages internationalizations.
+       * tests/netx/unit/net/sourceforge/jnlp/JNLPFileTest.java:
+       New unit test that checks the localesMatches method in JNLPFile.
+       * tests/netx/unit/net/sourceforge/jnlp/MockJNLPFile.java:
+       New class used to create a mock JNLPFile object.
+       * tests/netx/unit/net/sourceforge/jnlp/ParserTest.java:
+       New unit test that checks that the return of getTitle and getVendor
+       have localized information.

I would like to push this to HEAD and backport it to 1.3. The regression isn't present in 1.2 and before. Any thoughts?

[1] http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=955
-------------- next part --------------
A non-text attachment was scrubbed...
Name: title-vendor-required-06.patch
Type: text/x-patch
Size: 46673 bytes
Desc: not available
Url : http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20120801/3b914721/title-vendor-required-06.patch 

More information about the distro-pkg-dev mailing list