Bug 492838

Description Michael Vorburger 2016-05-02 13:13:15 EDT

This is a similar rant to my earlier bug 490001, but with an interesting twist:

Neon Milestone 6 (4.6.0M6) Build id: 20160329-0659 chokes on the following generics, in a project using Java 8 level, which works in javac because there is mvn CLI build which digs this, and IDEA according to other users:

this.substatements = ImmutableList.copyOf(Collections2.transform(substatementsToBuild, StmtContextUtils.buildEffective()));

saying: The method transform(Collection<F>, Function<? super F,T>) in the type Collections2 is not applicable for the arguments (Collection<StatementContextBase<?,?,?>>, Function<StmtContext<?,?,? extends EffectiveStatement<?,?>>,EffectiveStatement<?,?>>)

If interested, you should be to reproduce this in a complete WS quite easily, by using my https://github.com/vorburger/opendaylight-eclipse-setup and provisioning (just) the yangtools project.  Check out current git revision 31d5ed0c664e8fbd793538c4e4546204847c779e just to be sure we're talking about the same thing, and the error shown above will appear in line 94 of org.opendaylight.yangtools.yang-parser-impl/src/main/java/org/opendaylight/yangtools/yang/parser/stmt/rfc6020/effective/EffectiveStatementBase.java

Stephan I recall your "javac is known to have long standing bugs in this area, and reverse engineering these bugs will get us nowhere. The goal in implementing a Java compiler is to be compliant with JLS." from bug 490001, BUT here is the interesting twist on this one:

If you mark "StmtContextUtils.buildEffective()" and do a Refactor > Extract Local Variable, it will turn the above into this below, and on this it doesn't choke anymore:

Function<StmtContext<?, ?, ? extends EffectiveStatement<?, ?>>, EffectiveStatement<?, ?>> buildEffective = StmtContextUtils.buildEffective();
this.substatements = ImmutableList.copyOf(Collections2.transform(substatementsToBuild, buildEffective));

My point being: If Refactoring "knows" about the correct (horrible...) generic here, shouldn't & couldn't the type inference that's underlying the generics stuff be smart enough to also dig this, without requiring factoring out to an explicit variable ?

If there are other open bugs about this / something related to this, do link.  I had a look around but there are so many open generic related bugs (wow!) that it's hard to see what is what.


PS: I'll just be honest and just admit that I do actually understand this scary horrible looking generic - it's part of an API of a product I'm trying to get my head around (and haven't yet..).  From the little I understand, there are perhaps way too many <?, ?, ?> instead of <? extends ...> all over here? Thoughts welcome. Addressing this, project wide, is probably a major undertaking. Be that as it may, it would still seem "nice" if Eclipse would handle this code which javac & IDEA do.