Follow up 844

[tcharding]

This does everything in the issue description of #850. I'm not sure why I opened #856 also, and I'm confused by the first two check points of 856 as well.

Follow up #844.

Close #850

[tcharding]

Needs #860 to get past CI.

[apoelstra]

In 3f6c383:

You can't reexport the same symbol with different types based on a feature gate. This makes the features non-additive. Why export these?

[apoelstra]

f4d8fb4 looks good otherwise.

[tcharding]

This makes the features non-additive.

Oooph, of course.

[apoelstra]

ping @tcharding

[tcharding]

In 3f6c383:

You can't reexport the same symbol with different types based on a feature gate. This makes the features non-additive. Why export these?

AFAICT the type is set exactly by the feature std because of how we set the path to internal. So if std feature is on you get the std one irrespective of what other features are on. What am I missing?

#[cfg_attr(feature = "std", path = "internal_std.rs")]
#[cfg_attr(not(feature = "std"), path = "internal_nostd.rs")]
mod internal;

#[cfg(not(feature = "std"))]
mod spinlock;

pub use internal::{
    rerandomize_global_context, with_global_context, with_raw_global_context, SECP256K1,
};

[tcharding]

Rebased to see if lint warnings go away. No other changes.

[apoelstra]

So if std feature is on you get the std one irrespective of what other features are on. What am I missing?

If std is on you get a different type than if std is off. This means that std is not additive.

[tcharding]

oooo, I get it. You build with no-std, get one type, then someone else in your stack enables std and you get borked.