Error on Reserved Syntax for Raw Keywords: k#foo

[Julian-Wollersberger]

This PR makes the raw keyword syntax from rust-lang/rfcs#3098 a hard error on all editions. (In contrast to what the RFC currently says.)

error: raw keyword syntax is reserved for future use
   |
LL |     one_token!(k#ident);
   |                ^^^^^^^ help: insert a whitespace: `k #ident`

More specifically, it becomes a tokenization error, so macros cannot use it. k#ident is now lexed as one token, similar to r#ident, although this could be changed later without breaking anything.

Note that this is a breaking change, since macros previously interpreted it as three tokens, so something like quote!{ #k#other_variable } was valid. The crater run showed no real regressions though. You can easily get the old behavior by inserting a whitespace between k and #.

The actual raw keywords feature for things like k#async is not implemented here.

Previously, this PR was inteded to do a crater run for all reserved prefixes from rust-lang/rfcs#3101, but even the try build ran into breakage in the ecosystem, so only k#... was tested.

[rust-highfive]

r? @lcnr

(rust-highfive has picked a reviewer for you, use r? to override)

[joshtriplett]

@bors try

[bors]

⌛ Trying commit 692220c974656fac8908e4edc45b731a22fcb220 with merge 183585b6408d129d017caa116edd19c6d5f7f716...

[bors]

💔 Test failed - checks-actions

[Julian-Wollersberger]

I fixed the build errors in rustdoc, but the try build already ran into breakage in the ecosystem. tracing-attributes contains:

quote! {
    #name = #kind#value
}

Instances like this one should be somewhat common with the quote! macro. It's easily fixed though, by inserting a whitespace between #kind and #value.

What's the best way forward here? Should I just leave the error for raw keywords k#foo and remove the others? Or try to fix tracing-attributes?
AFAIK, only the raw keywords syntax was considered to use in older editions anyway.

[joshtriplett]

Yeah, let's prioritize getting a crater run that just checks if we can use k#keyword in every edition.

[Julian-Wollersberger]

Ok, I removed the other errors and updated the PR description.
@joshtriplett can you start another try run?

[tmiasko]

@bors try

[bors]

⌛ Trying commit 2bbdfdfac232fea427fdb311560411c39e0d6c39 with merge 2bf0c8e648350595527d984daafa09b33c338d13...

[bors]

☀️ Try build successful - checks-actions
Build commit: 2bf0c8e648350595527d984daafa09b33c338d13 (2bf0c8e648350595527d984daafa09b33c338d13)

[joshtriplett]

Looks like one test failed due to the change in lexer structures. Can you fix that up, please?

[Julian-Wollersberger]

Done.

[joshtriplett]

@bors try

[bors]

⌛ Trying commit 85844b1980f1137c308f83cae30bb5b174391c48 with merge b0d0ea232bdae12fa8da8491a91ae1db029990ef...

[bors]

☀️ Try build successful - checks-actions
Build commit: b0d0ea232bdae12fa8da8491a91ae1db029990ef (b0d0ea232bdae12fa8da8491a91ae1db029990ef)

[joshtriplett]

@craterbot check

[craterbot]

👌 Experiment pr-84037 created and queued.
🤖 Automatically detected try build b0d0ea232bdae12fa8da8491a91ae1db029990ef
⚠️ Try build based on commit a836d9b, but latest commit is 85844b1980f1137c308f83cae30bb5b174391c48. Did you forget to make a new try build?
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment pr-84037 is completed!
📊 6 regressed and 18 fixed (156232 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the blacklist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[joshtriplett]

I reviewed the crater results, and it looks like zero failures as a result of the k#foo reservation. So, I think we can safely reserve k#foo in all editions.

[Julian-Wollersberger]

I'll update this PR to more accurately implement the k#ident proposal, in the next week or so.

@rustbot label: -S-waiting-on-review +S-waiting-on-author

[petrochenkov]

r? @petrochenkov

[Julian-Wollersberger]

@petrochenkov this is ready for review now, and I guess also for lang team discussion. I also updated the PR description.
I redid the lexer changes, so it should accurately reflect the RFC (minus the r#$foo part). In essence, it does the same thing as r#ident now.

@scottmcm can you verify that this is what you had in mind in the RFC?

@rustbot label: +S-waiting-on-review -S-waiting-on-author

[petrochenkov]

@Julian-Wollersberger
Could you cooperate with @lrh2000 (#85359) and decide who is implementing what?
I didn't look at the changes in detail yet but these PRs appear to compete with each other.

[crlf0710]

Blocked on #85359 according to #85359 (comment)

[bors]

☔ The latest upstream changes (presumably #85359) made this pull request unmergeable. Please resolve the merge conflicts.

[Julian-Wollersberger]

This is still blocked on the RFC 3098, but work on that seems to have stalled.
I'll close this PR for now, but please ping me if the RFC is merged. I'm still interested in implementing it.