fixes #11225; generic sandwich problems; [backport:1.2]

[Araq]

No description provided.

[timotheecour]

• this example has many bugs (symbols should be exported, var O instead of O, wrong import, etc)
• after fixing those locally, the example works both before and after PR, so it's probably not the best example
• in future work, to avoid these problems (code block out of sync because not tested) we need HAR (see https://github.com/marler8997/har, I ported it to nim, see example har file here: https://github.com/timotheecour/vitanim/blob/master/testcases/har/test1.har) which can be serialized/deserialized to disk analog to tar, but human readable. some runnableExamples, tests, and some github issues would benefit from it whenever more than 1 test file are needed.
  (=> HAR: human readable archive timotheecour/Nim#637)

[timotheecour]

shouldn't nkMixinStmt be stripped away in semantic pass before it reaches backend, eg, so that it doesn't affect each backend

[Araq]

Maybe, but I'm generally in favor of not losing information in the AST. As you can see there are always things that backends explicitly need to ignore anyway.

[Araq]

Merging this for now as most changes are related to handle the nkMixinStmt, nkBindStmt in the backends which is worthwhile no matter what we end up with.

[mratsim]

So if I understand correctly the toSeq macro in sequtils should be patched to use bind items so that it stops being an annoyance with Hashsets, Tables, Intsets and all other collections?

[timotheecour]

• nit: toSeq is not a macro, it's (currently) 3 overloaded templates; this may affect how bind works
• I don't think this PR helps with for loop invoked from generic procedure defined in another module cant find items iterator #11167 if that's the issue you're referring to; the problem still holds after this PR with (adapted from for loop invoked from generic procedure defined in another module cant find items iterator #11167):

# a.nim
iterator items(a: int): int = discard
proc p*(a: int | uint) =
  # for _ in a.items: discard # ok
  bind items # doesn't help
  for _ in a: discard # still a bug

# b.nim
import a
p 1

still gives Error: type mismatch: got <int>

• see for loop invoked from generic procedure defined in another module cant find items iterator #11167 (comment) for directions to search for a bugfix

Regarding toSeq, I'm going to make the bold claim that iterable[T] from #17196 will fix all the problems with not just toSeq but many other issues related to called iterators being treated (currently) as untyped

[timotheecour]

potential root cause for #17704 regression