[Reverted] Improved annotations for select.select()

[nikhilm]

The TypeVars allow mypy to enforce that the returned lists are subsets of the arguments.

[gvanrossum]

Please fix up the indentation here.

[gvanrossum]

Thanks!

[gvanrossum]

Ow, this caused mypy's runtests.py to fail as follows:

PARALLEL 2
SUMMARY  139 tasks selected
passed 120, failed 1, pending 17; running 1                                     

FAILURE  #121 check stdlibsamples (3.2)

subprocess.py:1586: error: Need type annotation for variable
test/test_subprocess.py:1294: error: Need type annotation for variable

SUMMARY  1/139 tasks and 1/3981 tests failed
 FAILURE check stdlibsamples (3.2)
SUMMARY  1/139 tasks and 1/3981 tests failed
*** FAILURE ***

See discussion in #1059.

[gvanrossum]

So the specific example was something like

rlist, wlist, xlist = select(rlist, wlist, [])

so mypy now complains that it doesn't have an item type for xlist.

I'm not sure how to resolve this -- reverting this PR feels sad, but that might be a common case (since few people even know what to do with xlist).

[gvanrossum]

OTOH a simple solution would be to add casts to the sample code, e.g. select(rlist, wlist, cast(List, []). I'll try that.

[JelleZijlstra]

Hm, it seems like most code that uses select.select passes at least one empty list. I'd rather not force users to write a cast for each empty list argument, so maybe (unfortunately) it's better to revert.

[gvanrossum]

But we could at least preserve the arguments as Sequence[Any].

[gvanrossum]

BTW This is pretty common fallout from a sudden surge in contributions and nothing we should get too upset about. We just need to diligently fix the new issues.

[gvanrossum]

Yeah, I'm going to revert this. Please try again with just the argument types.

[nikhilm]

Thanks. will follow up in a few days.