build error: overflow evaluating the requirement F: FnMut<(&Pk,)>

[vincenzopalazzo]

I am building a project with bdk and I am now running the cargo install but
I had this error here

error[E0275]: overflow evaluating the requirement `F: FnMut<(&Pk,)>`
  |
  = help: consider increasing the recursion limit by adding a `#![recursion_limit = "256"]` attribute to your crate (`miniscript`)
  = note: required for `&mut F` to implement `FnMut<(&Pk,)>`
  = note: 128 redundant requirements hidden
  = note: required for `&mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut ...` to implement `FnMut<(&Pk,)>`
  = note: the full type name has been written to '/home/vincent/Github/lndart/lampo.rs/target/release/deps/miniscript-7895a7902d8f17ce.long-type-8047342695735668623.txt'

For more information about this error, try `rustc --explain E0275`.
error: could not compile `miniscript` (lib) due to previous error
warning: build failed, waiting for other jobs to finish...
error: failed to compile `lampod-cli v0.1.0 (/home/vincent/Github/lndart/lampo.rs/lampod-cli)`, intermediate artifacts can be found at `/home/vincent/Github/lndart/lampo.rs/target`
make: *** [Makefile:21: install] Error 101

I am not sure how to reproduce it, but I guess you could have some idea on what kind macros is causing this error?

[apoelstra]

Are you using a released version of Rust?

I believe you are hitting rust-lang/rust#110475 which should be fixed by #546.

If the bug is affecting released version of rustc then we need to backport the fix.

[vincenzopalazzo]

Ah, I am on nightly, but I guess I also tried with stable, and at this point, your fix should be included inside the bdk.

Thanks

[dpc]

Edit: Oh, now I see it was not backported to 8.0.1... . I updated to 9.0.0 since it seems to use the same bitcoin version.

I'm trying to update our CI to Rust 1.71 release. This forced me to update to a proper miniscript 8.0 release, which seems to work fine with existing 1.67.0 rustc.

> rustc --version
rustc 1.71.1 (eb26296b5 2023-08-03)

and i'm hitting:

fedimint-pkgs-deps>    Compiling miniscript v8.0.1
fedimint-pkgs-deps>    Compiling bitcoincore-rpc-json v0.16.0
fedimint-pkgs-deps>    Compiling electrum-client v0.12.1
fedimint-pkgs-deps>    Compiling esplora-client v0.5.0
fedimint-pkgs-deps>    Compiling secp256k1-zkp v0.7.0 (https://github.com/dpc/rust-secp256k1-zkp/?branch=sanket-pr#f29b1b8c)
fedimint-pkgs-deps> error[E0275]: overflow evaluating the requirement `F: FnMut<(&Pk,)>`
fedimint-pkgs-deps>   |
fedimint-pkgs-deps>   = help: consider increasing the recursion limit by adding a `#![recursion_limit = "256"]` attribute to your crate (`miniscript`)
fedimint-pkgs-deps>   = note: required for `&mut F` to implement `FnMut<(&Pk,)>`
fedimint-pkgs-deps>   = note: 128 redundant requirements hidden
fedimint-pkgs-deps>   = note: required for `&mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut &mut ...` to implement `FnMut<(&Pk,)>`
fedimint-pkgs-deps>   = note: the full type name has been written to '/build/source/target/release/deps/miniscript-ac0735e57f598110.long-type-2916159546684365784.txt'
fedimint-pkgs-deps> For more information about this error, try `rustc --explain E0275`.
fedimint-pkgs-deps> error: could not compile `miniscript` (lib) due to previous error
fedimint-pkgs-deps> warning: build failed, waiting for other jobs to finish...

The same code seems to fail with

[apoelstra]

Ok, I can backport to 8.0 as well.

What a freeking chore though. I'm tempted to just say "we don't support rustc after 1.71" because it's incredibly demoralizing to have to clean up after the compiler authors after giving them 3 months notice of a bug.

[dpc]

Since updating to 9.0 worked, I personally no longer need it. But technically 8.0 is a available, non-yanked version.

[apoelstra]

I'm not going to yank it. They should yank rustc.

But glad that it's no longer an urgent issue for you.

[dpc]

BTW. @apoelstra , @sanket1729 that &mut &mut &mut &mut doesn't look good. It seems to me like exactly same thing that was the rationale for rust-bitcoin/rust-bitcoin#1035 (comment) . It probably leads to large code generation increase.

On top of my head:

F: FnMut(&'a Pk) -> bool

would be better as:

F: &mut FnMut(&'a Pk) -> bool

Note: I might be wrong.

[apoelstra]

@dpc the rationale for rust-bitcoin 1030 was that we had multiple functions that would non-recursively call each other, resulting in a small finite number of &muts, which we were worried would lead to inefficient code generation (and in some cases, some functions would be recompiled with different numbers of &muts yes).

The consensus encoding traits in rust-bitcoin are also not really meant to be used directly, though they can be, so we were willing to make the API a bit worse to improve this.

The situation here is different -- these functions are only called from themselves, so they should be compiled only once (per choice of generics). Basically you either have zero &muts or infinity &muts, nothing in between.

With all of our other rust-miniscript APIs that have this issue, we've fixed by having a private internal function that takes a single &mut and a public one that doesn't.

[apoelstra]

I'm going to close this. We fixed it and backported to 4 or 5 old versions and nobody has complained about more versions in a while.