[joaodinissf]

AbstractSelectorFragmentProvider counted earlier siblings with the same selector value using equals() when computing a fragment, but getEObjectFromSegment matches them ignoring case unless the class is an ICaseSensitiveNamedElement. For a case-insensitive class with siblings foo and FOO, the fragment of FOO was …(…='FOO')#0, which resolved to foo. References, index entries and navigation to FOO silently landed on foo.

Both sides now use one matches helper.

Scope

• Only languages whose .export declares a non-unique uri-fragment = attribute(...) on a type that isn't an ICaseSensitiveNamedElement, with siblings whose values differ only in case (for example overloads in a case-insensitive language). DDK's own languages don't use uri-fragment.
• The only fragments that change are ones that previously resolved to the wrong object, so no correct stored fragment changes. Rebuilding affected resources refreshes the stale ones.
• The unique form is unchanged.

Tests. These are the first case-sensitivity tests in AbstractSelectorFragmentProviderTest, using a dynamic EMF model:

• a case-sensitive class distinguishes foo and FOO;
• a case-insensitive class resolves 'FOO' to foo;
• case-insensitive siblings foo and FOO each round-trip to themselves. This fails without the fix.

Follow-up to #1584.

🤖 Generated with Claude Code