No worries, @193s! Thank you for your contribution. It looks like the publish-core task discovered an issue with this submission:

Compiling 314 files to /home/jenkins/workspace/scala-2.11.x-validate-publish-core/build/locker/classes/compiler
/home/jenkins/workspace/scala-2.11.x-validate-publish-core/src/compiler/scala/tools/nsc/backend/jvm/BCodeBodyBuilder.scala:616: warning: matches with two cases or fewer are emitted using if-then-else instead of switch
              (argsSize : @switch) match {
                           ^
/home/jenkins/workspace/scala-2.11.x-validate-publish-core/src/compiler/scala/tools/reflect/WrappedProperties.scala:23: error: method propOrElse overrides nothing.
Note: the super classes of trait WrappedProperties contain the following, non final members named propOrElse:
def propOrElse(name: String,alt: => String): String
  override def propOrElse(name: String, alt: String) = wrap(super.propOrElse(name, alt)) getOrElse alt
               ^
/home/jenkins/workspace/scala-2.11.x-validate-publish-core/src/compiler/scala/tools/reflect/WrappedProperties.scala:26: error: method envOrElse overrides nothing.
Note: the super classes of trait WrappedProperties contain the following, non final members named envOrElse:
def envOrElse(name: String,alt: => String): String
  override def envOrElse(name: String, alt: String)  = wrap(super.envOrElse(name, alt)) getOrElse alt
               ^
one warning found
two errors found

I fixed! Thank you, @adriaanm!

BUILD SUCCESSFUL
Total time: 11 minutes 32 seconds

Great! You may ignore the IDE failure (#4287 has been merged to ignore it). The test failure is because this PR introduces a binary incompatible change in an API used by partest, which is compiled separately: scala/scala-partest.

Interestingly, test/files/run/t7775.scala relies on the eager behavior for the test to test anything. (Though I didn't test that.)

I wonder if a test rig could have the feature that it can demonstrate that it is still testing something.

A proof that the predicate being asserted is still meaningful after all these years.

edit (by lrytz): the test is extended in #4372

@adriaanm any ideas how we could push the change through (see your earlier comment)? i don't see a good way to keep the old signatures in the bytecode.

we could add overloads

propOrElse(name: String, alt: => String)
propOrElse(name: String, alt: String)

but then the compiler would pick the strict version, so the change would be pointless.

See #4297 for a similar problem.

@193s could you squash the three commits into one? (on your branch, run git rebase -i 70f960cfd1, then git push -f).

@lrytz I did as you said, thanks!

I think we can consider it a bug that the old *OrElse methods were eager, so we should fix that in 2.12. It seems unlikely that someone would expect eagerness here, or that these methods would be used in a performance hot spot. Let's add a deprecation message in 2.11 that says these methods will become call by name. The commit message of this PR should also explain why we made the change.

@adriaanm agreed, but my question was about the testing infrastructure: how can we get a green build, given that partest links against the old signature?

@193s thanks for the squash! as suggested by Adriaan, could you provide a bit more detail in the commit message? I created a ticket (https://issues.scala-lang.org/browse/SI-9246), so you can refer to that. See here for a lengthy description about commit messages: https://github.com/scala/scala/wiki/Pull-Request-Policy

@lrytz, @adriaanm Thanks for thanks for your kind advice.
I changed the commit message (e50fc2c), will that be ok?

@193s looks good, thanks! Now you just have to be patient until we figure out how to get our test infrastructure working.

We'll have to release a new version of partest that uses reflection for one minor release I guess. Or that doesn't use the methods that break bin compat.

I bet no one would notice if you sucked the partest module back into scala during lunch hour today. There wouldn't be even the slightest dip in the community contribution rate, either.

True, but I'd rather not eat lunch than re-ingest that module. Completely agreed the partest burden should not be on the contributor, though.

Hi, once #4419 is merged, please git rebase -i 2.11.x and git push --force 193s topic/prop2:topic/prop2. This should allow this PR to go green, since partest no longer uses the *OrElse methods.

I've opened a new PR that does the rebase. I'll close this one. Thanks again for your contribution & patience ;-)