Consider gating APIs that require winapi/advapi32/libc behind a feature

[LaylBongers]

Some crates (such as various vector math libraries) do not require things like cryptographically secure RNG. It might be beneficial to provide a way to use rand with these features disabled, even if they're enabled by default.

[Xaeroxe]

Might I ask, to what end? Having functions available in a crate that you don't use doesn't seem to hurt the final binary.

[aochagavia]

A concrete example: today I was trying to compile a program to WebAssembly, but it was impossible because rand relies on libc.

[dhardy]

That's a sound argument, if somewhat unfortunate (is there any proposal to use a fake libc when compiling for WASM?).

If we were to do this, it would be preferable to have the feature enabled by default (because frequently used functionality like thread_rng and random relies on it), but this could make it difficult linking rand without the feature (same as no_std: if any crate "forgets" to not depend on default features, Cargo will enable them).

I see you just added #197; is this enough to fix your problem or do you still have linker issues?

[aochagavia]

I just discovered an update to libc was published today, which provides compatibility with wasm. That, together with my PR allows things to compile, though I haven't tested the resulting code on the browser.

[pitdicker]

I don't think it is helpful to keep this issue open. The original issue about gating APIs that require linking behind a feature gate is solved now that we have the std feature. Only OsRng required linking to WinAPI, libc etc.

[dhardy]

Yes, we have native WASM support now. If rand won't compile on a specific platform, please open an issue regarding that platform, not a general one.

[huangjj27]

it seems compiled in yew will cause failure. Is it caused by the thread_rng function ?

[vks]

This would be a breaking change.

…

On Wed, May 9, 2018, 17:36 Junjie Huang ***@***.***> wrote: it seems compiled in yew will cause failure <yewstack/yew#227>. Is it caused by the thread_rng function ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#89 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AACCtH2wHxMJFZMtGZsvZH7c_jAIdqw4ks5tww0PgaJpZM4G79ia> .

[pitdicker]

This would be a breaking change.

Be careful to know the state of the issue when replying from e-mail 😄.