Rumor has it that there is too much queries to SourceForBinaryQuery class during opening of project, computing its
classpath transitive closure and parsing it. Please measure how much time/disk access can be eliminated by caching
queries for the same JARs/folders and implement an intelligent cache that will invalidate itself when query result
The rumors did not mention, that the problem is perhaps only related API support projects. I tried it with several
simple J2SE projects and was unable to reproduce it on J2SE. But - just to be sure - we need to setup more project with
complex dependencies to ensure that J2SE is not affected.
We discussed possible solution with Jarda and agreed that I will try to transform several API support modules with their
dependencies to J2SE projects.
We agreed we will test it in maven projects too.
Fix of #162706 improved behavior for projects with many dependencies.