-
Notifications
You must be signed in to change notification settings - Fork 12.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
expand: Stop using nonterminals for passing tokens to attribute and derive macros #73345
Conversation
@bors try |
⌛ Trying commit afc5180669a14fa5d65f86e5f65f96976e927bd5 with merge c8bdaa84ec1b37f18f30375e196a65b7602f3790... |
☀️ Try build successful - checks-azure |
@craterbot check |
👌 Experiment ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more |
🚧 Experiment ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more |
🎉 Experiment
|
All randomly selected regressions that I've analyzed are caused by some popular proc macro crate using an assert checking that It's very easy to fix, but I still need to triage all the regressions to find out other possible root causes. |
Summary of errors:
From those error at least this subset
is caused by these asserts in https://github.com/servo/rust-cssparser/blob/master/procedural-masquerade/lib.rs#L201-L232 I'll try to hard-code all this stuff in the compiler to avoid regressions until the proper solution with keeping token jointness is implemented. |
I’ve reproduced this error by running this in an older commit the
I can publish a new version of Compiling cssparser v0.26.0 (/home/simon/projects/rust-cssparser)
thread '<unnamed>' panicked at 'expected suffix "," not found in "# [allow(unused)] enum ProceduralMasqueradeDummyType\n{\n Input =\n (0, stringify !\n (\"deg\" => v, \"grad\" => v * 360. / 400., \"rad\" => v * 360. / (2. * PI),\n \"turn\" => v * 360., _ => return\n Err(location .\n new_unexpected_token_error(Token :: Ident(unit . clone()))),)) . 0\n}"', procedural-masquerade/lib.rs:233:9
[…]
error: proc-macro derive panicked I can hack it some more to support both with and without the comma, but it gets annoying for a crate that at this point only exists to support old compilers between Rust 1.15 and 1.29. (Those with stable proc macro derive but not other proc macros.) diff --git procedural-masquerade/lib.rs procedural-masquerade/lib.rs
index 2736ccf..c1a99a1 100644
--- procedural-masquerade/lib.rs
+++ procedural-masquerade/lib.rs
@@ -199,14 +199,23 @@ pub fn _extract_input(derive_input: &str) -> &str {
let mut input = derive_input;
for expected in &[
- "#[allow(unused)]",
+ "#",
+ "[",
+ "allow",
+ "(",
+ "unused",
+ ")",
+ "]",
"enum",
"ProceduralMasqueradeDummyType",
"{",
"Input",
"=",
- "(0,",
- "stringify!",
+ "(",
+ "0",
+ ",",
+ "stringify",
+ "!",
"(",
] {
input = input.trim_start();
@@ -219,7 +228,7 @@ pub fn _extract_input(derive_input: &str) -> &str {
input = &input[expected.len()..];
}
- for expected in [")", ").0,", "}"].iter().rev() {
+ for expected in [")", ")", ".", "0", ",", "}"].iter().rev() {
input = input.trim_end();
assert!(
input.ends_with(expected), |
Oh but if I add the trailing comma to the original dummy enum then it’s preserved. I’ve submitted servo/rust-cssparser#274 |
@SimonSapin |
Blocked on merging servo/rust-cssparser#274 (UPD: Done) and publishing a bugfix version of The asserts checking for whitespace can be fixed up in the compiler, but the assert checking for trailing comma unfortunately can not. |
Make procedural-masquerade resilient to macro input pretty-printing changes CC rust-lang/rust#73345 (comment)
https://crates.io/crates/procedural-masquerade/0.1.7 is up and should be picked up as a recursive dependency of some of those crates that failed the last Crater run. |
expand: Stop using nonterminals for passing tokens to attribute and derive macros Make one more step towards fully token-based expansion and fix issues described in rust-lang#72545 (comment). Now `struct S;` is passed to `foo!(struct S;)` and `#[foo] struct S;` in the same way - as a token stream `struct S ;`, rather than a single non-terminal token `NtItem` which is then broken into parts later. The cost is making pretty-printing of token streams less pretty. Some of the pretty-printing regressions will be recovered by keeping jointness with each token, which we will need to do anyway. Unfortunately, this is not exactly the same thing as rust-lang#73102. One more observable effect is how `$crate` is printed in the attribute input. Inside `NtItem` was printed as `crate` or `that_crate`, now as a part of a token stream it's printed as `$crate` (there are good reasons for these differences, see rust-lang#62393 and related PRs). This may break old proc macros (custom derives) written before the main portion of the proc macro API (macros 1.2) was stabilized, those macros did `input.to_string()` and reparsed the result, now that result can contain `$crate` which cannot be reparsed. So, I think we should do this regardless, but we need to run crater first. r? @Aaron1011
ast_pretty: Pass some token streams and trees by reference Salvaged from an intermediate version of rust-lang#73345.
ast_pretty: Pass some token streams and trees by reference Salvaged from an intermediate version of rust-lang#73345.
…arth Rollup of 16 pull requests Successful merges: - rust-lang#72569 (Remove legacy InnoSetup GUI installer) - rust-lang#73306 (Don't implement Fn* traits for #[target_feature] functions) - rust-lang#73345 (expand: Stop using nonterminals for passing tokens to attribute and derive macros) - rust-lang#73449 (Provide more information on duplicate lang item error.) - rust-lang#73569 (Handle `macro_rules!` tokens consistently across crates) - rust-lang#73803 (Recover extra trailing angle brackets in struct definition) - rust-lang#73839 (Split and expand nonstandard-style lints unicode unit test.) - rust-lang#73841 (Remove defunct `-Z print-region-graph`) - rust-lang#73848 (Fix markdown rendering in librustc_lexer docs) - rust-lang#73865 (Fix Zulip topic format) - rust-lang#73892 (Clean up E0712 explanation) - rust-lang#73898 (remove duplicate test for rust-lang#61935) - rust-lang#73906 (Add missing backtick in `ty_error_with_message`) - rust-lang#73909 (`#[deny(unsafe_op_in_unsafe_fn)]` in libstd/fs.rs) - rust-lang#73910 (Rewrite a few manual index loops with while-let) - rust-lang#73929 (Fix comment typo) Failed merges: r? @ghost
We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work.
We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work.
Convert `proc_macro_back_compat` lint to an unconditional error. We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work. r? `@estebank`
…stebank Convert `proc_macro_back_compat` lint to an unconditional error. We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work. r? `@estebank`
…stebank Convert `proc_macro_back_compat` lint to an unconditional error. We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work. r? ``@estebank``
Rollup merge of rust-lang#125596 - nnethercote:rental-hard-error, r=estebank Convert `proc_macro_back_compat` lint to an unconditional error. We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work. r? ``@estebank``
We still check for the `rental`/`allsorts-rental` crates. But now if they are detected we just emit a fatal error, instead of emitting a warning and providing alternative behaviour. The original "hack" implementing alternative behaviour was added in rust-lang#73345. The lint was added in rust-lang#83127. The tracking issue is rust-lang#83125. The direct motivation for the change is that providing the alternative behaviour is interfering with rust-lang#125174 and follow-on work.
Make one more step towards fully token-based expansion and fix issues described in #72545 (comment).
Now
struct S;
is passed tofoo!(struct S;)
and#[foo] struct S;
in the same way - as a token streamstruct S ;
, rather than a single non-terminal tokenNtItem
which is then broken into parts later.The cost is making pretty-printing of token streams less pretty.
Some of the pretty-printing regressions will be recovered by keeping jointness with each token, which we will need to do anyway.
Unfortunately, this is not exactly the same thing as #73102.
One more observable effect is how
$crate
is printed in the attribute input.Inside
NtItem
was printed ascrate
orthat_crate
, now as a part of a token stream it's printed as$crate
(there are good reasons for these differences, see #62393 and related PRs).This may break old proc macros (custom derives) written before the main portion of the proc macro API (macros 1.2) was stabilized, those macros did
input.to_string()
and reparsed the result, now that result can contain$crate
which cannot be reparsed.So, I think we should do this regardless, but we need to run crater first.
r? @Aaron1011