Regression: unsafe block needed

[gilescope]

My compiler suggests that the unsafe blocks around the unsafe unreachable functions are mandatory.

They were removed recently in PR: #9637 so this is just putting them back.

[ordian]

hmm, this function is marked as safe in the latest nightly https://doc.rust-lang.org/nightly/core/arch/wasm32/fn.unreachable.html
and since we're compiling to wasm using nightly, it should be fine, no?

jumped the gun

[gilescope]

Jan points out that it compiles with latest nightly: https://doc.rust-lang.org/nightly/core/arch/wasm32/fn.unreachable.html

[bkchr]

Please update your compiler.

This code is also compiled by CI

[gilescope]

Do we want to only support the latest nightly compiler? We could keep the unsafe and have a warning suppression attribute in there, and that would increase the range of nightly compilers that people could use (with a comment to remove the suppression once it's 2022).

At the moment this change is force upgrading everyone for not much reason.

[bkchr]

Do we want to only support the latest nightly compiler?

Yes

[bkchr]

No need to support any old nightlies.

[chevdor]

This is an issue I also see in srtool to build all the runtimes. Without the fix, we would have to drop using stable (1.53.0) and go 'back' to nightlies. I tested also with stable 1.55.0 without success.

[koute]

This is an issue I also see in srtool to build all the runtimes. Without the fix, we would have to drop using stable (1.53.0) and go 'back' to nightlies. I tested also with stable 1.55.0 without success.

FYI, the unsafe is also unnecessary in the current beta, which should be promoted to stable next month on 2021-10-21.

[s3krit]

I'm having an issue again with running the benchmarks against westend, kusama and polkadot for Polkadot v0.9.12 due to this issue - that is based on the substrate polkadot-v0.9.12 branch, which is simply branched from master a few days ago. It can be seen here: https://gitlab.parity.io/parity/polkadot/-/jobs/1181692 - what are my options?

• Should I apply the change in this PR as a hotfix against the release-v0.9.12
• Should I wait until this issue is resolved in stable? Thus holding up the release of v0.9.12 for some amount of time?
• Should we be using nightly rather than stable for building? If so, I'd need some assistance ASAP on how to do this to unblock the v0.9.12 branch.

My preference right now would be the first option and simply apply this hotfix to the polkadot-v0.9.12 branch so that we're getting builds working again and the release is unblocked, but I'm a bit out of my field of expertise here.

CC @bkchr maybe you can help? :)

[bkchr]

@s3krit Yes add the commit to the release branch. We should make CI build everything with stable in the near future.