Pants tests induce PEXWarning from resolve package collisions

[cburroughs]

Pants tests induce PEXWarning from resolve package collisions

[cburroughs]

A default of False in 2.20 and then True in 2.21 to de-risk 2.20 makes sense to me.

Pants is already pretty chatty so I do lean towards True in 2.21 even if every single case isn't squashed. This issue with typing_extensions is "probably fine" and could trivially be worked around by futzing with the pants internal lockfiles, but the same warning could be telling a user something closer "Your code is broken!" and I don't like swallowing the underlying tool trying to tell them that.

[cburroughs]

Although now that I've thought through that a little more, we probably should keep it enabled for Pants itself (in pants.toml) so we keep surfacing new cases as dogfood.

[huonw]

Good catch!

I would think it'd be good not to ship this and #20577 in a stable release, potentially we should have the new option default to False for 2.20.x (and adjust the docs to indicate that there's a few "spurious" issues)?

This would relieve the pressure of needing to land potentially-broad changes to fix the underlying issues (which we presumably should fix) to unblock 2.20.0.

Thoughts?

[huonw]

(Chucked up #20590 with that idea, in case we want to go with it.)

[huonw]

Removing this from the 2.20.x milestone, since we landed #20593 as a stop-gap to reduce severity in 2.20.0