I18628

[EnzeXing]

This PR fixes the minimized test of #18628 that previously runs infinitely on the global initialization checker.

[liufengyun]

LGTM, thanks a lot.

Please add the long test as a positive test (no warnings).

[EnzeXing]

Please add the long test as a positive test (no warnings).

@liufengyun With this fix I'm afraid the long test cannot pass the checker since the environment of repMany will eventually widen the by-name parameters to Cold

[liufengyun]

Please add the long test as a positive test (no warnings).

@liufengyun With this fix I'm afraid the long test cannot pass the checker since the environment of repMany will eventually widen the by-name parameters to Cold

It should work --- I checked on my machine, it works as expected. The case is OK, because the recursive call is by-name.

[olhotak]

I confirm that the long test also works for me. (It gives a warning but does not run out of memory). Please add it, but otherwise LGTM.

[EnzeXing]

I confirm that the long test also works for me. (It gives a warning but does not run out of memory). Please add it, but otherwise LGTM.

It is the same for me, but @liufengyun do you expect the test to pass without warning?

[liufengyun]

I confirm that the long test also works for me. (It gives a warning but does not run out of memory). Please add it, but otherwise LGTM.

It is the same for me, but @liufengyun do you expect the test to pass without warning?

Sorry, I made a mistake in my test. Yes, it produce warnings. But it shouldn't --- It seems we have a problem handling the local lazy vals.

Just change the following method in the test from

 def ~ [U](q: => Parser[U]): Parser[~[T, U]] = { lazy val p = q // lazy argument (for(a <- this; b <- p) yield new ~(a,b)) }

to

 def ~ [U](q: => Parser[U]): Parser[~[T, U]] = { (for(a <- this; b <- q) yield new ~(a,b)) }

will suppress the warning.