[3.8] bpo-38122: minor fixes to AsyncMock spec handling (GH-16099).

[matrixise]

(cherry picked from commit 14fd925)

Co-authored-by: Michael Foord voidspace@users.noreply.github.com

https://bugs.python.org/issue38122

[matrixise]

Hi @voidspace

There is problem with the #14700 because that one has not been backported to 3. and your code use _AnyComparer. I try to find the right solution with @pganssle

[matrixise]

@voidspace _AnyComparer is specific to master, not backported to 3.8, so, I have removed that class from your PR and run the tests. wait & see.

[lisroach]

LGTM

[tirkarthi]

Seems this missed the RC build :( It more looks like a code refactor. If I understand it correctly in previous case there was list comprehension over arguments starting with "spec" to take first argument and perhaps depending on the order they are passed. Now with PR it checks for spec_set and falls back to spec making the choice predictable to detect. I guess it could cause problem if someone depends on the order in 3.8 RC or 3.8.0 and upgrades to 3.8.1 as it's merged later. Also there is a removed comment about what to do when spec_set and spec are different so could it be tested to ensure this choice is explicit?

[matrixise]

I worked on this issue in september but 3.8 has been released in october, do you think we could merge it into 3.8?

[csabella]

@matrixise, should this change still be backported to 3.8? Thanks!

[matrixise]

@csabella for this PR, yep, I think we should backport it to 3.8.

[ambv]

Sorry, closing this stale backport as 3.8 is closing to security fixes only mode.