Couldn't resolve reference to JvmIdentifiableElement '==' [message #1118458] |
Fri, 27 September 2013 08:32  |
Eclipse User |
|
|
|
Hello gurus, I'm hitting a Couldn't resolve reference to JvmIdentifiableElement '=='..
Any tips where exactly this (validation) error is created and how to start debugging this?
FYI We're hitting this while trying to move from Xtext 2.3.1 (all working) to 2.4.2 (currently broken) in https://github.com/vorburger/eLang/commits/v2.2_xtext-v2.4.2 .. the ch.vorburger.el.engine.tests.BasicExpressionGeneratorTest.testNotNull() illustrates this problem (if anybody does try; ignore other test failures related to problems in Interpreter Tests which broke as well following the update, they're less critical).
I've seen a number of posts re. this all saying "missing xbase.lib!", but that doesn't seem to be it (xbase.lib IS on ch.vorburger.el)... must be something more... interesting.
Any help much appreciated.
|
|
|
|
|
Re: Couldn't resolve reference to JvmIdentifiableElement '==' [message #1122677 is a reply to message #1121430] |
Tue, 01 October 2013 16:44  |
Eclipse User |
|
|
|
Managed to figure this out; just to wrap this up FTR, in case anybody hits something like this (no response needed, unless some you have some further insight so that I can learn something):
It seemed to have had something to do with org.eclipse.xtext.resource.XtextResource.RESOLVE_ALL=true when loading the (synthetic) resource.. I don't fully understand it, but it's reproducible: If you checkout commit fd5eed335ab51f0e3889815c7b426f1b005cc5f7 of https://github.com/vorburger/eLang, the BasicExpressionGeneratorTest works. If you now go into the ExpressionFactory and in line 109 pass resourceSet.getLoadOptions() instead of null, then testNotNull fails with Couldn't resolve reference to JvmIdentifiableElement '=='.
I've tried to provide a reproducible standalone example test case of this in https://github.com/vorburger/xtext-sandbox/tree/f63a74c328afb3a56ffd45deb07b0d5012c051d1/XbaseOptionResolveAllConfusion, but there what I thought was basically the same does work... so it must be the side effect of something else.
I learnt that I also had to write a *JvmModelInferrer, even though previously was not needed and I had assumed for the UC in Q here that was not needed.
[Updated on: Tue, 01 October 2013 16:44] by Moderator
|
|
|
Powered by
FUDForum. Page generated in 0.22191 seconds