inference: fix fieldtype tfunc design

[vtjnash]

fix #31666 and add some tests for nothrow computation

[oschulz]

@ararslan , @JeffBezanson : Could this be backported to the upcoming v1.1.1 as well? Or is it too late for that?

[JeffBezanson]

Should be safe to backport, along with #31848.

[KristofferC]

1.1.1 is too far in the release process to add anything to.

[ararslan]

Yes, 1.1.1 is frozen; it is effectively finished and will be tagged within a few days.

[oschulz]

Sure, I understand, no worries.

[oschulz]

Would be the a backport candidate for the next Julia LTS release, when the time comes (1.0.x or 1.1.x)?

[ararslan]

There likely won't be any more 1.1.x releases, but the change will be in 1.2.0. (Indeed it should already be present in the release candidate.) As for 1.0.x, I see no reason not to backport it.

[oschulz]

Thanks for the info, @ararslan! An eventual backport to 1.0.x would be nice (and I get that 1.1 can't be maintained long term, with 1.2 and 1.3 in the pipeline).

Oh, and yes, we had no more trouble with this since v1.2-rc2.