Breaking change re: Error implementations on Windows

[pie-flavor]

The re-addition of the default-implementations of std::error::Error without the std feature got all the relevant Linux platforms, but missed Windows, making it still a breaking change. I'm experiencing this issue downstream from bcrypt, which no longer compiles on Windows without overriding the getrandom dep.

[newpavlov]

I think it should be fixed on the bcrypt side by forwarding the std feature. Plus simultaneously they also can update to getrandom v0.2, which does require such forwarding on Linux targets as well.

[josephlr]

Looks like this was resolved in Keats/rust-bcrypt#57. However, we should still probably update the 0.1 branch to fix this. It was a breaking change, and adding the std stuff back for Windows shouldn't hurt anything.

[newpavlov]

IIRC std stuff never worked on Windows in the first place, so it's not a breaking change, but an inconsistency between Linux and Windows targets.

[josephlr]

IIRC std stuff never worked on Windows in the first place, so it's not a breaking change, but an inconsistency between Linux and Windows targets.

You're right.

I confirmed that v0.1.8 never had the std::error::Error implementations for Windows (unless you had the "std" feature enabled). So the removal and readding were not breaking changes on Windows. Closing as not a bug.

[pie-flavor]

@josephlr 0.1.6 did, though. That just means the break took place slightly earlier.

[josephlr]

@josephlr 0.1.6 did, though. That just means the break took place slightly earlier.

Good catch, reopening.