[1.38] Placement new syntax no longer parses

[Mark-Simulacrum]

root: mbox 0.3.1; cc @kennytm causes:

• resin-os/healthdog-rs: start v. end; cc @willnewton

root: rayon-hash 0.2.0 cc @cuviper

• coder543/xray: start v. end; cc @coder543

Root cause of these on our side is #60803 which removed support for parsing placement new even under #[cfg(false)] and the like. The crater run done there seems to have not found these -- not sure why. We might be testing more now than we were before.

[kennytm]

The <- support in mbox has been removed in 0.5.0.

The mbox 0.3.1 dependency was introduced by systemd 0.1.0, but the current version is systemd 0.4.0 which doesn't depend on mbox.

resin-os/healthdog-rs depends on systemd 0.* so a cargo update is enough to remove the broken mbox dependency.

FYI @willnewton you may want to run cargo update to update the Cargo.lock.

[cuviper]

root: rayon-hash 0.2.0 cc @cuviper

• coder543/xray: start v. end; cc @coder543

rayon-hash came up in #50832 and was fixed in 0.3.0. It's up to 0.5.0 now, and deprecated in favor of hashbrown/rayon anyway. @coder543 should update, if this xray project is at all active.

[coder543]

@cuviper xray is unfortunately not being developed any further at this time

[coder543]

@Mark-Simulacrum those "ccs" you tried to do to tag people don't work because you put them in code blocks, just FYI...

[cuviper]

@coder543 would you accept a PR bumping dependencies? Otherwise, maybe we should remove this repo from crater's list.

[coder543]

@cuviper yeah, either option would be fine!

[Mark-Simulacrum]

@Mark-Simulacrum those "ccs" you tried to do to tag people don't work because you put them in code blocks, just FYI...

That's intentional, I try to not ping downstream users before the root cause is assessed and/or addressed by that author.

[coder543]

Updated code has been merged

[nagisa]

Since this syntax was never proposed for stablisation, I think it is fine to say it is an oversight/bug this unstable syntax was not gated.

If we decide we cannot feasibly to keep unstable syntax properly gated (because it would not work with cfg, then <- parsing should be restored.

[nagisa]

We should decide how we want to deal with unstable syntax.

If unstable syntax like in this particular instance the <- operator, was properly gated, people would have to always use nightly if any code uses features with unstable syntax, even if such uses appeared behind #[cfg]. The current situation ended up not gating the syntax properly which is now causing problems when changing/removing it. Tagging T-lang to make them aware of it.