Imported From: https://issues.scala-lang.org/browse/SI-8627?orig=1
Reporter: Chris Howitt (chowitt)
Affected Versions: 2.11.0, 2.11.1, 2.11.4
See #3265, #8552

Chris Howitt (chowitt) said:
Looks like this has been broken and fixed before:
#6440
scala/scala@7a6905d
There's even a test: test/files/run/t6440.scala

Here is the commit that broke this again:
scala/scala@3059e3a#diff-4b8694484188574dce55135fd0bc9ae5

@ruippeixotog said:
I have just issued a pull request for this:
scala/scala#3925

@gkossakowski said:
Fixed in scala/scala#3949

Venkatesh Umaashankar (vumaasha) said:
it looks like this is again broken in 2.11.4, I am trying this in scala REPL and it does not work. I am not sure if I should log a new bug?

@gkossakowski said:
The change that fixed this ticket has been reverted in: scala/scala#4048

We forgot to reopen the corresponding issue.

@Ichoran said:
@lrytz Did you want to fix this one, or shall I grab it?

@lrytz said:
@Ichoran I'm Happy to hand it off :-)

@Ichoran said:
@lrytz I guess I should go for the ugly hack that fixes it in 2.11.5 rather than waiting for 2.12? Or shall we just say this one's buggered 'til then?

@lrytz said:
I'd also go for the fix in 2.11.5.

@adriaanm said:
AFAICT, still pending because the original fix was not binary compatible. Should we retarget to 2.12?

@Ichoran said:
@adriaanm I need to do more extensive testing to see if we will clobber performance too greatly in 2.11 if we fix this the way that will work with binary compatibility (i.e. intercept it in the superclass). Since this is a potentially dangerous change, I want to make sure to test in a variety of scenarios (e.g. with and without multiple dispatch). I will retarget if the results of testing are disappointing.

@Ichoran said:
@adriaanm Not fixable in 2.11 without significant performance penalty (up to about 40% slowdown in List.filterNot on single-element lists). Deferring to 2.12. How can we document this so that the lack of laziness of filterNot is discoverable in 2.11?

@Ichoran said:
[~lrytz] Your previous fix (already applied in 2.12) looks good modulo leaving the stub override of filter in Stream. When I figure out whether I can fix the docs without overriding anything I'll add a small additional patch. This should resolve this ticket. (I may also collapse the separate function--not really a compelling reason for it any more.)

@cvogt said:
reproduced in 2.11.4

@cvogt said:
This is a regression. There should be a test Stream.from(1).filterNot(_ <= 5)

@Ichoran said:
@cvogt The fix version is 2.12.0-M1. 2.11.everything is going to be affected because, as explained above, it cannot be fixed unless we either break binary compatibility or introduce major performance penalties on other collections.

And yes, of course, the offending line in the bug report will go in a test.

@adriaanm said:
scala/scala#4429

Espen Wiborg (espenhw) said:
AFAICT the referenced pull request does not fix this issue; instead it fixes #8552

@Ichoran said:
[~espenhw] Thanks for catching that! I've updated the tickets accordingly.

@SethTisue said:
This appears to work in 2.12.0-M1, presumably thanks to scala/scala#4284

@SethTisue said:
test coverage: scala/scala#4611

@lrytz said (edited on Jul 16, 2015 9:05:54 AM UTC):
i think the commit that fixed this ticked in 2.12.x is scala/scala#3949 (which was merged forward into 2.12.x). The commit was later reverted in 2.11.x due to binary incompatibility (scala/scala#4048), but the revert was annotated [nomerge], which means it was merged into 2.12.x with -s ours.