I've released an unofficial patch set, and would really appreciate feedback and testing. The patch sets are available all-in-one here:
They can be applied against the official build 1405 source.
If you'd like a drop-in replacement I've built a jar available here:
To read more, touch this text and see your path into the rabbit hole...
Fix duplicate Javadoc entries: javadoc-duplicate.patch Each Javadoc entry was generated twice. This fixes that.
Add navbar and page CSS attributes: fproxy-attributes.patch As requested by FreeStyle (/Freetalk/showThread?BoardName=eng.freenet&ThreadID=776332e1-4c3c-478e-bc7d-5cc063a3c989@4nZEgeYC3x2HXdhYGvSnuF~MxrddmNNrKs~SCX2gGIE)
The current path is now appended to "page-" in an id on the <body>. The fproxy-page attribute on the <body> has been changed from an id to a class. None of the official themes reference this attribute. Fred navlist entries now have an id based on the localization key. Plugin navlist entries now have an id based on the localization key appended to the class name, separated by a dash. The navlist-selected attribute has been changed from an id to a class. The official themes are updated to reflect this. Disk browser fixes file-browser-cleanup.patch Removes state from the browser so reentrantcy or getting stuck should not be a problem. Adds null checks to hopefully avoid NPEs. Node permissions should now be used correctly - previously the check was for upload permission in all cases. Directories that are not allowed are not rendered. This means that if allowed directories are disjoint that they cannot be browsed between. A virtual top-level directory to allow access to these would work, but is not implemented. Alert configuration changes fix useralert-config.patch Alerts that change configuration values were missing the attribute to specify which subconfig the desired value was part of. Now alerts that ask for a configuration change should work as intended again. Queue toadlet cleanup queue-cleanup.patch Refactoring to replace an enum pattern with a real enum and verbose global l10n() calls with a local equivalent. No functional changes. Please let me know what you think, or if problems turn up. I hope to continue releasing patches similarly, so if there's anything I can do to present this better please mention it.