Disable bindgen and upgrade libsecp256k1

[xoloki]

Fixes #82

[cloudflare-workers-and-pages]

Deploying p256k1 with    Cloudflare Pages

Latest commit:	83fa426
Status:	✅  Deploy successful!
Preview URL:	https://ea1450f6.p256k1.pages.dev
Branch Preview URL:	https://disable-bindgen-and-upgrade.p256k1.pages.dev

View logs

[jcnelson]

I can confirm that both cargo build and cargo test succeed on musl libc Linux without any special tweaking.

[djordon]

I take it that this is basically just a git pull && git checkout TAG from the libsecp256k1 repo? Or is it more than just that?

Side note: I would consider using git submodules for the secp256k1 source. I assume everything in the p256k1/_secp256k1 folder is basically a specific commit of the secp256k1 repo, but it's quite hard to tell for sure. Using a submodule would make it clear what's going on and make it easy to point out any of the changes on top of the source from secp256k1. And if you want to ensure the source of the submodule is always under your control, I'd set up a mirror of the bitcoin-core/secp256k1 repo and point my submodule to that. The main benefit here is that the diffs are super clear and easy to reconcile. Any thoughts?

[xoloki]

I take it that this is basically just a git pull && git checkout TAG from the libsecp256k1 repo? Or is it more than just that?

It's a download not a checkout:

https://github.com/Trust-Machines/p256k1/blob/master/update/src/main.rs#L11

Side note: I would consider using git submodules for the secp256k1 source. I assume everything in the p256k1/_secp256k1 folder is basically a specific commit of the secp256k1 repo, but it's quite hard to tell for sure. Using a submodule would make it clear what's going on and make it easy to point out any of the changes on top of the source from secp256k1. And if you want to ensure the source of the submodule is always under your control, I'd set up a mirror of the bitcoin-core/secp256k1 repo and point my submodule to that. The main benefit here is that the diffs are super clear and easy to reconcile. Any thoughts?

That's how I originally structured the repo. But in bf6d61b it was changed to do the current download/patch/vendor/define dance. Part of the reason was to get rid of conflicting symbols when stacks-core had multiple versions of different secp256k1 wrappers. But yeah definitely worth reconsidering now that bindgen is no longer working as it did originally.