Singleton types not inferred in all places they should be even after r25861

[scabug]

With Scala trunk, following r25861 and with -Yexperimental specified.

In the REPL transcript below, the compiler infers T as the non-singleton type String, and then rejects it as not conforming to the explicitly provided bound <: Singleton.

miles@lewis:scala$ scala-trunk -Xexperimental
Welcome to Scala version 2.10.0.r25866-b20111021020217 (Java HotSpot(TM) 64-Bit Server VM, Java 1.7.0).
Type in expressions to have them evaluated.
Type :help for more information.

scala> val x = "one"
x: String = one

scala> def single[T <: Singleton](x : T) = x
single: [T <: Singleton](x: T)T

scala> single(x)
<console>:10: error: inferred type arguments [String] do not conform to method single's type parameter bounds [T <: Singleton]
              single(x)
              ^

scala> single[x.type](x)
res2: x.type = one

[scabug]

Imported From: https://issues.scala-lang.org/browse/SI-5103?orig=1
Reporter: @milessabin

[scabug]

@adriaanm said:
Unassigning and rescheduling to M6 as previous deadline was missed.

[scabug]

@retronym said:
I had a quick stab at implementing this. I can get these tests to pass, but couldn't stop other tests from failing.

https://github.com/retronym/scala/compare/ticket/5103

/cc [~folone]

[scabug]

@milessabin said:
I've given this a quick spin and it covers a lot of my cases ... I would very likely be able to remove the singleton type related macros from shapeless with this fix :-)

[scabug]

@milessabin said:
Where are we with this getting merged? Any chance of having it in 2.11.2?

[scabug]

@retronym said:
It's not ready for the primetime, unfortunately.

I got stuck and couldn't figure out how to take the step forward without breaking existing code.

From the commit comment:

Checking in again with `pos` reveals breakage around typetags and
manifests, among others.

test/partest --update-check
/Users/jason/code/scala/test/files/pos/annotated-treecopy
/Users/jason/code/scala/test/files/pos/macro-implicit-invalidate-on-error.scala
/Users/jason/code/scala/test/files/pos/manifest1-new.scala
/Users/jason/code/scala/test/files/pos/nothing_manifest_disambig-new.scala
/Users/jason/code/scala/test/files/pos/nothing_manifest_disambig-old.scala
/Users/jason/code/scala/test/files/pos/t4501.scala
/Users/jason/code/scala/test/files/pos/t5022.scala
/Users/jason/code/scala/test/files/pos/t7190.scala
/Users/jason/code/scala/test/files/pos/t7516
/Users/jason/code/scala/test/files/pos/t7649.scala
/Users/jason/code/scala/test/files/pos/t7818.scala

```

[scabug]

@milessabin said:
What sort of breakage are you seeing? I vaguely remember that there was quite a bit of unnecessary T <: Singleton sprinkled over tags and manifests when plain old dependent method types would do ... are these the things falling foul of the fix?

[scabug]

@SethTisue said:
@milessabin does your SIP-23 work affect this?

[scabug]

@milessabin said:
Yes, this is fixed on that branch. Initial work by Adriaan with some tweaks subsequently by me.

[scabug]

@SethTisue said:
haha I'd better not forget that with Adriaan sitting next right to me! :-)

scala/scala#5310

[scabug]

@milessabin said:
Why has this been put back to 2.12.2?

[scabug]

@adriaanm said:
I was making a sweep through tickets assigned to 2.12.1 to get them down to a reasonable number given the deadline of Nov 30.