SWC - Update to 2.3.x causes incorrect output about bitflags HOT 9 CLOSED

Thanks for the report @MolotovCherry. I’ll check this out. I wouldn’t expect things to break between minor releases unless there’s some reliance on specific debug formatting.

from bitflags.

I found the root cause of the bug, and it's not an issue with bitflags, rather swc doing something unsanctioned that just happened to work for about 5 years until it didn't because of a change in implementation detail.

I'll write up the issue (or a PR) with details on the swc repo and link here.

from bitflags.

The original bug here should be fixed as of 2.3.3.

from bitflags.

Thanks for the update @kosayoda. Those are my favourite kinds of bugs 😄

I’ve been on holidays this week so haven’t had a chance to dig into this yet.

from bitflags.

Following the linked PR it looks like you might have run into this issue.

For what it’s worth, I think that is a bug in bitflags and am currently working through revamping our test suite to better specify the behaviour of methods when there are bits set that don’t correspond to known flags.

from bitflags.

Ah that does make sense. I think it boils down to whether adding new flags external to the bitflags macro ("external API" use case in the linked issue) should be sanctioned, which on first instinct I don't think it should due to the amount of edge cases in the implementation details.

from bitflags.

I think where possible we’ll try respect bits that don’t correspond to known flags. So if the bitwise operation on the underlying integer would retain that bit then the operation on the flags type will too.

The only case that gets a bit weird I think is !, where we don’t want to set all unset bits, so it just unconditionally clears any bits that don’t correspond to known flags.

from bitflags.

I've opened #366 to fix this, and think we've hashed out a plan for a 3.x release that should make these operators fully consistent in their support flags types with externally defined flags.

If you've got any thoughts on that I'd love to hear them!

from bitflags.

This is great. I tested 2.3.3 with the original example which had the problem, and it now shows the correct output. Thanks a lot!

from bitflags.