Original Yrecursion tests progressed

[SethTisue]

The test case in question dates back to 2008. Without -Yrecursion 1, It fails to compile in Scala 2.7.7, but succeeds in Scala 2.8.2 and other subsequent Scala 2 versions.

(On Scala 3 it doesn't compile at all because it uses # on an abstract type, which the language no longer permits.)

A pull request removing -Yrecursion 1 from the test would be welcome, afaics. Adding a new test case that does require -Yrecursion 1 in order to compile would be welcome as well.

[SethTisue]

Why(!) is -Yrecursion not needed anymore for this test-case?

We don't know — and probably never will. It was so long ago that doing the necessary archaeology to answer that would be rather difficult.

We routinely use git bisect to identify the particular PR when something progressed or regressed, if we care enough about the issue in question, but we can only reasonably (as in, without substantial extra effort) go back a certain number of years that way, and Scala 2.7 and 2.8 are now outside of that window.

[som-snytt]

That's a good one. I've wanted a partest option to "fuzz" test options. Sometimes I'll write a test with -Xlint but it's not part of the test, only to lint the test code.

[som-snytt]

other history ancient and modern which demonstrates the long haul of cyclic aliasing. https://github.com/scala/bug/issues/8252

showing sometimes there is a reply from on high that does not answer our prayers https://github.com/scala/bug/issues/11963

(Edit: the latter was an illustration for any concerned interlopers of a question going unanswered [because there is no answer] and a polite appeal to powers that be, which elicited a response of a sort.)