Use new --output-format=doctest rustdoc command line flag to improve doctest handling.

Thanks for the detailed answer!

I added the missing "Signed-of-by" in the commit message.

I added some more context in the commit message, hopefully I didn't
miss anything.

> We currently support versions 1.78+, so we will need to do these
> things conditionally. I can help with that, so don't worry too much
> about it for the moment (e.g. we have `rustc-option` and other helpers
> to pick the right flags given a version etc.).

I'll definitely need some help here. I'm not sure current stable already
has this change so until then, we'll need a beta/nightly version to run
these tests.

> Hmm... I was hoping we could do something on the `rustdoc` side to
> remove these hacks altogether since we are going to have a "proper
> flag" for it, i.e. to avoid relying on the particular "format" of the
> tests somehow.

I opened rust-lang/rust#137807 to resolve
this problem.

> > +    let doctest_code = doctest_code.replace(
> > +        "} _inner().unwrap() }",
> > +        "} let test_return_value = _inner(); assert!(test_return_value.is_ok()); }",
> >      );
> > +    std::fs::write(path, doctest_code.as_bytes()).unwrap();
> > +}
>
> Same for this.

For this one I'll need to check first if it can be done "safely" (ie,
so unexpected side
effect).

> > +    } else {
> > +        panic!("missing `format_version` field");
> > +    }
>
> `expect` maybe?

I don't think `expect` would work in any of the cases in this file.
What I suggest
is to add methods on `JsonValue` in a future patch which would allow to reduce
code in this file (and call `expect` too).

> These two bits could go in a first patch, I think, though it isn't a big deal.

You're absolutely right, removing them and sending a separate patch.

Here is the updated patch:

The goal of this patch is to remove the use of 2 unstable rustdoc features
(`--no-run` and `--test-builder`) and replace it with a stable feature:
`--output-format=doctest`, which was added in
rust-lang/rust#134531.

Before this patch, the code was using very hacky methods in order to retrieve
doctests, modify them as needed and then concatenate all of them in one file.

Now, with this new flag, it instead asks rustdoc to provide the doctests
code with their associated information such as file path and line number.

Signed-off-by: Guillaume Gomez <[email protected]>

Author: GuillaumeGomez

rust/Makefile

@@ -171,14 +171,14 @@ quiet_cmd_rustdoc_test_kernel = RUSTDOC TK $<
 	rm -rf $(objtree)/$(obj)/test/doctests/kernel; \
 	mkdir -p $(objtree)/$(obj)/test/doctests/kernel; \
 	OBJTREE=$(abspath $(objtree)) \
-	$(RUSTDOC) --test $(rust_flags) \
+	$(RUSTDOC) --output-format=doctest $(rust_flags) \
 		-L$(objtree)/$(obj) --extern ffi --extern kernel \
 		--extern build_error --extern macros \
 		--extern bindings --extern uapi \
-		--no-run --crate-name kernel -Zunstable-options \
+		--crate-name kernel -Zunstable-options \
 		--sysroot=/dev/null \
-		--test-builder $(objtree)/scripts/rustdoc_test_builder \
-		$< $(rustdoc_test_kernel_quiet); \
+		$< $(rustdoc_test_kernel_quiet) > rustdoc.json; \
+	cat rustdoc.json | $(objtree)/scripts/rustdoc_test_builder; \
 	$(objtree)/scripts/rustdoc_test_gen
 
 %/doctests_kernel_generated.rs %/doctests_kernel_generated_kunit.c: \

scripts/json.rs

@@ -0,0 +1,235 @@
+// SPDX-License-Identifier: GPL-2.0
+
+//! JSON parser used to parse rustdoc output when retrieving doctests.
+
+use std::collections::HashMap;
+use std::iter::Peekable;
+use std::str::FromStr;
+
+#[derive(Debug, PartialEq, Eq)]
+pub(crate) enum JsonValue {
+    Object(HashMap<String, JsonValue>),
+    String(String),
+    Number(i32),
+    Bool(bool),
+    Array(Vec<JsonValue>),
+    Null,
+}
+
+fn parse_ident<I: Iterator<Item = char>>(
+    iter: &mut I,
+    output: JsonValue,
+    ident: &str,
+) -> Result<JsonValue, String> {
+    let mut ident_iter = ident.chars().skip(1);
+
+    loop {
+        let i = ident_iter.next();
+        if i.is_none() {
+            return Ok(output);
+        }
+        let c = iter.next();
+        if i != c {
+            if let Some(c) = c {
+                return Err(format!("Unexpected character `{c}` when parsing `{ident}`"));
+            }
+            return Err(format!("Missing character when parsing `{ident}`"));
+        }
+    }
+}
+
+fn parse_string<I: Iterator<Item = char>>(iter: &mut I) -> Result<JsonValue, String> {
+    let mut out = String::new();
+
+    while let Some(c) = iter.next() {
+        match c {
+            '\\' => {
+                let Some(c) = iter.next() else { break };
+                match c {
+                    '"' | '\\' | '/' => out.push(c),
+                    'b' => out.push(char::from(0x8u8)),
+                    'f' => out.push(char::from(0xCu8)),
+                    't' => out.push('\t'),
+                    'r' => out.push('\r'),
+                    'n' => out.push('\n'),
+                    _ => {
+                        // This code doesn't handle codepoints so we put the string content as is.
+                        out.push('\\');
+                        out.push(c);
+                    }
+                }
+            }
+            '"' => {
+                return Ok(JsonValue::String(out));
+            }
+            _ => out.push(c),
+        }
+    }
+    Err(format!("Unclosed JSON string `{out}`"))
+}
+
+fn parse_number<I: Iterator<Item = char>>(
+    iter: &mut Peekable<I>,
+    digit: char,
+) -> Result<JsonValue, String> {
+    let mut nb = String::new();
+
+    nb.push(digit);
+    loop {
+        // We peek next character to prevent taking it from the iterator in case it's a comma.
+        if matches!(iter.peek(), Some(',' | '}' | ']')) {
+            break;
+        }
+        let Some(c) = iter.next() else { break };
+        if c.is_whitespace() {
+            break;
+        } else if !c.is_ascii_digit() {
+            return Err(format!("Error when parsing number `{nb}`: found `{c}`"));
+        }
+        nb.push(c);
+    }
+    i32::from_str(&nb)
+        .map(|nb| JsonValue::Number(nb))
+        .map_err(|error| format!("Invalid number: `{error}`"))
+}
+
+fn parse_array<I: Iterator<Item = char>>(iter: &mut Peekable<I>) -> Result<JsonValue, String> {
+    let mut values = Vec::new();
+
+    'main: loop {
+        let Some(c) = iter.next() else {
+            return Err("Unclosed array".to_string());
+        };
+        if c.is_whitespace() {
+            continue;
+        } else if c == ']' {
+            break;
+        }
+        values.push(parse(iter, c)?);
+        while let Some(c) = iter.next() {
+            if c.is_whitespace() {
+                continue;
+            } else if c == ',' {
+                break;
+            } else if c == ']' {
+                break 'main;
+            } else {
+                return Err(format!("Unexpected `{c}` when parsing array"));
+            }
+        }
+    }
+    Ok(JsonValue::Array(values))
+}
+
+fn parse_object<I: Iterator<Item = char>>(iter: &mut Peekable<I>) -> Result<JsonValue, String> {
+    let mut values = HashMap::new();
+
+    'main: loop {
+        let Some(c) = iter.next() else {
+            return Err("Unclosed object".to_string());
+        };
+        let key;
+        if c.is_whitespace() {
+            continue;
+        } else if c == '"' {
+            let JsonValue::String(k) = parse_string(iter)? else {
+                unreachable!()
+            };
+            key = k;
+        } else if c == '}' {
+            break;
+        } else {
+            return Err(format!("Expected `\"` when parsing Object, found `{c}`"));
+        }
+
+        // We then get the `:` separator.
+        loop {
+            let Some(c) = iter.next() else {
+                return Err(format!("Missing value after key `{key}`"));
+            };
+            if c.is_whitespace() {
+                continue;
+            } else if c == ':' {
+                break;
+            } else {
+                return Err(format!(
+                    "Expected `:` after key, found `{c}` when parsing object"
+                ));
+            }
+        }
+        // Then the value.
+        let value = loop {
+            let Some(c) = iter.next() else {
+                return Err(format!("Missing value after key `{key}`"));
+            };
+            if c.is_whitespace() {
+                continue;
+            } else {
+                break parse(iter, c)?;
+            }
+        };
+
+        if values.contains_key(&key) {
+            return Err(format!("Duplicated key `{key}`"));
+        }
+        values.insert(key, value);
+
+        while let Some(c) = iter.next() {
+            if c.is_whitespace() {
+                continue;
+            } else if c == ',' {
+                break;
+            } else if c == '}' {
+                break 'main;
+            } else {
+                return Err(format!("Unexpected `{c}` when parsing array"));
+            }
+        }
+    }
+    Ok(JsonValue::Object(values))
+}
+
+fn parse<I: Iterator<Item = char>>(iter: &mut Peekable<I>, c: char) -> Result<JsonValue, String> {
+    match c {
+        '{' => parse_object(iter),
+        '"' => parse_string(iter),
+        '[' => parse_array(iter),
+        't' => parse_ident(iter, JsonValue::Bool(true), "true"),
+        'f' => parse_ident(iter, JsonValue::Bool(false), "false"),
+        'n' => parse_ident(iter, JsonValue::Null, "null"),
+        c => {
+            if c.is_ascii_digit() || c == '-' {
+                parse_number(iter, c)
+            } else {
+                Err(format!("Unexpected `{c}` character"))
+            }
+        }
+    }
+}
+
+impl JsonValue {
+    pub(crate) fn parse(input: &str) -> Result<Self, String> {
+        let mut iter = input.chars().peekable();
+        let mut value = None;
+
+        while let Some(c) = iter.next() {
+            if c.is_whitespace() {
+                continue;
+            }
+            value = Some(parse(&mut iter, c)?);
+            break;
+        }
+        while let Some(c) = iter.next() {
+            if c.is_whitespace() {
+                continue;
+            } else {
+                return Err(format!("Unexpected character `{c}` after content"));
+            }
+        }
+        if let Some(value) = value {
+            Ok(value)
+        } else {
+            Err("Empty content".to_string())
+        }
+    }
+}

scripts/rustdoc_test_builder.rs

@@ -15,58 +15,93 @@
 //! from that. For the moment, we generate ourselves a new name, `{file}_{number}` instead, in
 //! the `gen` script (done there since we need to be aware of all the tests in a given file).
 
+use std::fs::create_dir_all;
 use std::io::Read;
 
-fn main() {
-    let mut stdin = std::io::stdin().lock();
-    let mut body = String::new();
-    stdin.read_to_string(&mut body).unwrap();
+use json::JsonValue;
 
-    // Find the generated function name looking for the inner function inside `main()`.
-    //
-    // The line we are looking for looks like one of the following:
-    //
-    // ```
-    // fn main() { #[allow(non_snake_case)] fn _doctest_main_rust_kernel_file_rs_28_0() {
-    // fn main() { #[allow(non_snake_case)] fn _doctest_main_rust_kernel_file_rs_37_0() -> Result<(), impl core::fmt::Debug> {
-    // ```
-    //
-    // It should be unlikely that doctest code matches such lines (when code is formatted properly).
-    let rustdoc_function_name = body
-        .lines()
-        .find_map(|line| {
-            Some(
-                line.split_once("fn main() {")?
-                    .1
-                    .split_once("fn ")?
-                    .1
-                    .split_once("()")?
-                    .0,
-            )
-            .filter(|x| x.chars().all(|c| c.is_alphanumeric() || c == '_'))
-        })
-        .expect("No test function found in `rustdoc`'s output.");
+mod json;
 
-    // Qualify `Result` to avoid the collision with our own `Result` coming from the prelude.
-    let body = body.replace(
-        &format!("{rustdoc_function_name}() -> Result<(), impl core::fmt::Debug> {{"),
-        &format!("{rustdoc_function_name}() -> core::result::Result<(), impl core::fmt::Debug> {{"),
-    );
+fn generate_doctest(file: &str, line: i32, doctest_code: &str) {
+    let file = file
+        .strip_suffix(".rs")
+        .unwrap_or(file)
+        .strip_prefix("../rust/kernel/")
+        .unwrap_or(file)
+        .replace('/', "_");
+    let path = format!("rust/test/doctests/kernel/{file}-{line}.rs");
 
+    // We replace the `Result` if needed.
+    let doctest_code = doctest_code.replace(
+        "fn main() { fn _inner() -> Result<",
+        "fn main() { fn _inner() -> core::result::Result<",
+    );
     // For tests that get generated with `Result`, like above, `rustdoc` generates an `unwrap()` on
     // the return value to check there were no returned errors. Instead, we use our assert macro
     // since we want to just fail the test, not panic the kernel.
     //
     // We save the result in a variable so that the failed assertion message looks nicer.
-    let body = body.replace(
-        &format!("}} {rustdoc_function_name}().unwrap() }}"),
-        &format!("}} let test_return_value = {rustdoc_function_name}(); assert!(test_return_value.is_ok()); }}"),
+    let doctest_code = doctest_code.replace(
+        "} _inner().unwrap() }",
+        "} let test_return_value = _inner(); assert!(test_return_value.is_ok()); }",
     );
+    std::fs::write(path, doctest_code.as_bytes()).unwrap();
+}
+
+fn main() {
+    let mut stdin = std::io::stdin().lock();
+    let mut body = String::new();
+    stdin.read_to_string(&mut body).unwrap();
+
+    let JsonValue::Object(rustdoc) = JsonValue::parse(&body).unwrap() else {
+        panic!("Expected an object")
+    };
+    if let Some(JsonValue::Number(format_version)) = rustdoc.get("format_version") {
+        if *format_version != 1 {
+            panic!("unsupported rustdoc format version: {format_version}");
+        }
+    } else {
+        panic!("missing `format_version` field");
+    }
+    let Some(JsonValue::Array(doctests)) = rustdoc.get("doctests") else {
+        panic!("`doctests` field is missing or has the wrong type");
+    };
+
+    // We ignore the error since it will fail when generating doctests below if the folder doesn't
+    // exist.
+    let _ = create_dir_all("rust/test/doctests/kernel");
 
-    // Figure out a smaller test name based on the generated function name.
-    let name = rustdoc_function_name.split_once("_rust_kernel_").unwrap().1;
+    let mut nb_generated = 0;
+    for doctest in doctests {
+        let JsonValue::Object(doctest) = doctest else {
+            unreachable!()
+        };
 
-    let path = format!("rust/test/doctests/kernel/{name}");
+        // We check if we need to skip this test by checking it's a rust code and it's not ignored.
+        if let Some(JsonValue::Object(attributes)) = doctest.get("doctest_attributes") {
+            if attributes.get("rust") != Some(&JsonValue::Bool(true)) {
+                continue;
+            } else if let Some(JsonValue::String(ignore)) = attributes.get("ignore") {
+                if ignore != "None" {
+                    continue;
+                }
+            }
+        }
+        if let (
+            Some(JsonValue::String(file)),
+            Some(JsonValue::Number(line)),
+            Some(JsonValue::String(doctest_code)),
+        ) = (
+            doctest.get("file"),
+            doctest.get("line"),
+            doctest.get("doctest_code"),
+        ) {
+            generate_doctest(file, *line, doctest_code);
+            nb_generated += 1;
+        }
+    }
 
-    std::fs::write(path, body.as_bytes()).unwrap();
+    if nb_generated == 0 {
+        panic!("No test function found in `rustdoc`'s output.");
+    }
 }