Rebased PR on top of master to hopefully fix spurious CI failures.

The macOS build keeps failing in a test which seems to be comparing some translate-c output, but I'm not sure how to even read/diagnose the build output. Seems to be completely unrelated to this PR though.

It's the pinned issue: #6830

Been trying to figure it out all day, sorry about that.

It's the pinned issue: #6830

Been trying to figure it out all day, sorry about that.

Thanks for the heads up!

Note to self: Read the actual pinned issues when something fishy seems to be going on! 😄

Rebased PR on top of master once more now that #6830 is fixed.

This was fixed concurrently by @ifreund in 1b34365; rebased/updated my PR to make the error unreachable.

This was fixed concurrently by @ifreund in 1b34365; rebased/updated my PR to make the error unreachable.

IMO leaking the abstraction of std.os and making it unreachable isn't the way to go. perhaps the errorset could exclude that error when not using windows?

IMO leaking the abstraction of std.os and making it unreachable isn't the way to go. perhaps the errorset could exclude that error when not using windows?

Since the #6600 proposal to get rid of/rework std.os has been accepted, perhaps we should defer this decision and decide this (how error sets deal with OS-specific errors) for the new high-level OS abstraction as a whole?

Since the #6600 proposal to get rid of/rework std.os has been accepted, perhaps we should defer this decision and decide this (how error sets deal with OS-specific errors) for the new high-level OS abstraction as a whole?

I think your change is a nice improvement in the meantime :)