Pin proc-macro2 in CI to fix MSRV breakage

[tnull]

The proc-macro2 crate switched to Rust edition 2021 starting with v1.0.66, i.e., has MSRV of 1.56.

Here, we pin it in CI to fix the breakage.

[TheBlueMatt]

Also, this is a pretty big dep tree just for tokio::select!() on one line, especially when we already have a two-select implemented in lightning-background-processor. I wonder if we shouldn't just copy that into lightning-net-tokio.

[tnull]

Also, this is a pretty big dep tree just for tokio::select!() on one line, especially when we already have a two-select implemented in lightning-background-processor. I wonder if we shouldn't just copy that into lightning-net-tokio.

Yeah, maybe. On the other hand, as discussed recently, a slight MSRV bump might be useful/imminent anyways, which would not only spare us to copy and maintain more code, but also would allow us to remove a lot (if not all) of the CI pins. Will go through what exactly our requirements are in the coming days to hopefully arrive at a sane target.

[TheBlueMatt]

On the other hand, as discussed recently, a slight MSRV bump might be useful/imminent anyways

I'm not sure why - the furthest we could realistically go is 1.63, and sadly there's really just not much interesting between 1.48 and 1.63...we'll probably have some other sub-crates with MSRVs of 1.63, but there's no reason to rush ahead and bump the MSRV of the other crates for some time, I think.

[tnull]

On the other hand, as discussed recently, a slight MSRV bump might be useful/imminent anyways

I'm not sure why - the furthest we could realistically go is 1.63, and sadly there's really just not much interesting between 1.48 and 1.63...we'll probably have some other sub-crates with MSRVs of 1.63, but there's no reason to rush ahead and bump the MSRV of the other crates for some time, I think.

Well, the transaction-sync crate MSRV should lie somewhere around ~1.56/1.57 with TLS currently? So, while it won't protect us from future MSRV bumps longer term, it would allow us to get started, until we can move to a saner dependecy tree there. IIRC, this 'two-pronged' approach was what we arrived at?

[TheBlueMatt]

Well, the transaction-sync crate MSRV should lie somewhere around ~1.56/1.57 with TLS currently?

Right, but there's no reason that has to imply a workspace-wide MSRV bump, the rest of the workspace can still happily sit at 1.48.

[TheBlueMatt]

I think we should copy the one from lightning-background-processor - its like 15 LoC (would become 20 with return elements, maybe) and removes several crates from our dependency tree.

[dunxen]

I feel like dropping the dependency is ideal here, but also happy with a temporary pin.

[tnull]

I feel like dropping the dependency is ideal here, but also happy with a temporary pin.

Yeah, we probably should. Will pick it up tomorrow, if @TheBlueMatt doesn't beat me to it. Closing this for now.