Community
Participate
Working Groups
To solve the bug #128866, we are required to change the version of the plugin.xml file to be <?eclipse version="3.2"?> instead of <?eclipse version="3.0"?>. This change is not implemented yet. When it will be some tweak will be required to PDE. Here is the list of things I can think of: - when <?eclipse version="3.0"?> is used report warnings when an extension id contains "." - when <?eclipse version="3.2"?> is used no warning is required when an extension id contains "." - probably change the way PDE computes fully qualified names for extension / extension points to properly handles the eclipse version markup - change the wizards to generate <?eclipse version="3.2"?> when the plug-in is for eclipse 3.2 and <?eclipse version="3.0"?> before that.
done. PDE now generates and handles 3.2 processing instructions. Incidentally, since anything seems to go as far as ID syntax goes, PDE is staying out of the warning business if an ID contains a dot.
Are you doing any verification if the eclipse tag is set to 3.0?
not sure what you mean in comment 2. verification of what? that an ID does not contain a dot if the processing version is 3.0? If so, then as I said in comment 1, no. PDE is staying out of that business since the runtime handles it fine.
actually we do it (and have always done it) for extension point IDs, but not for extension IDs. An extension ID syntax varies per extension point. We don't validate it, since the schema for an extension point just says that it's a string.
>An extension ID syntax varies per extension point. We don't validate it, since >the schema for an extension point just says that it's a string. The whole issue we encountered with M5 was caused by the fact that the extension ID is not just a string and is a simpleToken as defined in the plugin.dtd (see http://help.eclipse.org/help31/index.jsp?topic=/org.eclipse.platform.doc.isv/reference/misc/plugin_manifest.html) and that PDE use to be too permissive. ExtensionId := SimpleToken SimpleToken := sequence of characters from ('a-z','A-Z','0-9','_')
The plugin.dtd also says that it's optional, but the Core extension points (e.g. buiders, natures, products, ..) make it 'required' in their schemas. So the plugin.dtd behaviour can be overridden evidently. It is too late to legislate the use and syntax of this ID, particularly when the runtime tolerates both forms whether or not the processing instruction is 3.0 or 3.2.
