Added WithdrawWithheldTokensFromAccounts IX

[deanmlittle]

It seems WithdrawWithheldTokensFromAccounts ix was missing from token22 transfer fee extensions, so I implemented it.

You can check out the instruction here

It allows the fee withdraw authority to withdraw fees from t22 transfer fee extension accounts directly without having to first harvest to mint, then harvest from mint. Code is tested and working locally.

This unlocks the ability for Anchor programs to delegate a PDA signer as the token withdraw authority, drastically improving security, as any signer can invoke an IX to claim tokens to a treasury without having to first claim to a mint account.

[vercel]

@deanmlittle is attempting to deploy a commit to the coral-xyz Team on Vercel.

A member of the Team first needs to authorize it.

[acheroncrypto]

Thanks, could you note this feature in the CHANGELOG?

[deanmlittle]

Added a Changelog entry. FYI: I forked the v0.30.1 tag instead of master. Not sure if this is a standard way of making a PR, but the new discriminator changes in master were causing backwards compatibility issues with the CLI, so I wanted to isolate for that.

[acheroncrypto]

FYI: I forked the v0.30.1 tag instead of master.

It would be better to fork the branch you're PRing to (master in this case), otherwise CI won't work as expected.

the new discriminator changes in master were causing backwards compatibility issues with the CLI

Could you expand on this? I don't think the new custom discriminator changes should cause any compatibility issues for the CLI — in fact the CLI code hasn't changed at all to support it (14cec14...c5337c5).

There was a discriminator-related breakage from v0.29 -> v0.30 because of the new IDL spec (#2824) and required discriminator field, but there shouldn't be any backwards compatibility issue in CLI or IDL between v0.30.1 and latest.

[deanmlittle]

If you import the current master into an anchor program using git in the Cargo toml and try to run anchor build using CLI 0.30.1, idl-build will fail due to a lifetime issue on the DISCRIMINATOR and recommend you change the lifetime to 'static.

[acheroncrypto]

Thanks, the reason for this could be that you haven't changed anchor-spl to git version, or there might be another dependency that depends on an older version of Anchor.

The CLI itself is unaware of the &'static [u8] type change introduced in #3098, as it just builds and runs the code coming from the codegen.

There will be a conflict in the history due to the CHANGELOG override, could you rebase this branch to master?

[deanmlittle]

I do not believe that to be the case. I do not have any third-party packages, and am importing both lang and spl from my own branch forked from master which also implements the added ix above. There seems to be a genuine backwards compatibility issue there.

As this branch is currently being used in prod due the ix not being available in anchor yet, if you would like me to submit a version forked from master, I will have to create a new PR from that other branch that I initially forked from master, with the caveat that although the code would be identical, I cannot make the claim that it is tested and working locally with CLI v0.30.1 due to the aforementioned discriminator lifetime backwards-compatibility issue.

[deanmlittle]

Found the culprit: https://github.com/coral-xyz/anchor/blob/master/spl/src/idl_build.rs#L13

If we replace &[u8] with &'static[u8] it builds with 0.30.1.

If you would like me to make a combination of these two fixes with this PR, I can do that.

[acheroncrypto]

Found the culprit: https://github.com/coral-xyz/anchor/blob/master/spl/src/idl_build.rs#L13

If we replace &[u8] with &'static[u8] it builds with 0.30.1.

Hm, what's your nightly version? The recent versions the nightly compiler, which we use to build the IDL, should resolve elided lifetimes of associated constants to static if no other lifetimes are in scope. See rust-lang/rust#125190

Repro:

trait Discriminator {
    const DISCRIMINATOR: &'static [u8];
}

struct MyStruct;

impl Discriminator for MyStruct {
    const DISCRIMINATOR: &[u8] = &[];
}

cargo b

fails with rust-lang/rust#115010 (stable), but

cargo +nightly b

compiles.

If you would like me to make a combination of these two fixes with this PR, I can do that.

It wouldn't hurt to annotate the static lifetime, but since the changes are unrelated, it would be better if you could create a separate PR for the lifetime change.

[deanmlittle]

Will try nightly and report back shortly. Your explanation makes sense. Sounds like it is a good idea to make that PR anyway so will do that first then rebase this from updated master.

[deanmlittle]

Will rebase from here here once merged

[deanmlittle]

I have rebased to current master if you wanted to merge this first. The two shouldn't conflict anyway.

[acheroncrypto]

Thank you!