Rollup merge of #141004 - matthewjasper:unicode-before-expansion, r=davidtwco

Report text_direction_codepoint_in_literal when parsing

The lint is now reported in code that gets removed/modified/duplicated by macro expansion, and spans are more accurate so we don't get ICEs from trying to split a span in the middle of a character.

This removes support for lint level attributes for `text_direction_codepoint_in_literal` except at the crate level, I don't think that there's an easy way around this when the lint can be reported on code that's removed by `cfg` or that is only in the input of a macro.

Fixes #140281

Author: matthiaskrgr

Cargo.lock

@@ -2171,6 +2171,7 @@ dependencies = [
 name = "lint-docs"
 version = "0.1.0"
 dependencies = [
+ "rustc-literal-escaper",
  "serde_json",
  "tempfile",
  "walkdir",

compiler/rustc_lint/src/early/diagnostics.rs

@@ -187,6 +187,27 @@ pub fn decorate_builtin_lint(
                 lints::ReservedMultihash { suggestion }.decorate_lint(diag);
             }
         }
+        BuiltinLintDiag::HiddenUnicodeCodepoints {
+            label,
+            count,
+            span_label,
+            labels,
+            escape,
+            spans,
+        } => {
+            lints::HiddenUnicodeCodepointsDiag {
+                label: &label,
+                count,
+                span_label,
+                labels: labels.map(|spans| lints::HiddenUnicodeCodepointsDiagLabels { spans }),
+                sub: if escape {
+                    lints::HiddenUnicodeCodepointsDiagSub::Escape { spans }
+                } else {
+                    lints::HiddenUnicodeCodepointsDiagSub::NoEscape { spans }
+                },
+            }
+            .decorate_lint(diag);
+        }
         BuiltinLintDiag::UnusedBuiltinAttribute { attr_name, macro_name, invoc_span } => {
             lints::UnusedBuiltinAttribute { invoc_span, attr_name, macro_name }.decorate_lint(diag);
         }

compiler/rustc_lint/src/hidden_unicode_codepoints.rs

@@ -48,7 +48,6 @@ mod errors;
 mod expect;
 mod for_loops_over_fallibles;
 mod foreign_modules;
-pub mod hidden_unicode_codepoints;
 mod if_let_rescope;
 mod impl_trait_overcaptures;
 mod internal;
@@ -92,7 +91,6 @@ use deref_into_dyn_supertrait::*;
 use drop_forget_useless::*;
 use enum_intrinsics_non_enums::EnumIntrinsicsNonEnums;
 use for_loops_over_fallibles::*;
-use hidden_unicode_codepoints::*;
 use if_let_rescope::IfLetRescope;
 use impl_trait_overcaptures::ImplTraitOvercaptures;
 use internal::*;
@@ -177,7 +175,6 @@ early_lint_methods!(
             DeprecatedAttr: DeprecatedAttr::default(),
             WhileTrue: WhileTrue,
             NonAsciiIdents: NonAsciiIdents,
-            HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
             IncompleteInternalFeatures: IncompleteInternalFeatures,
             RedundantSemicolons: RedundantSemicolons,
             UnusedDocComment: UnusedDocComment,

compiler/rustc_lint/src/lib.rs

@@ -103,6 +103,7 @@ declare_lint_pass! {
         TAIL_EXPR_DROP_ORDER,
         TEST_UNSTABLE_LINT,
         TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
+        TEXT_DIRECTION_CODEPOINT_IN_LITERAL,
         TRIVIAL_CASTS,
         TRIVIAL_NUMERIC_CASTS,
         TYVAR_BEHIND_RAW_POINTER,
@@ -3782,7 +3783,6 @@ declare_lint! {
 }
 
 declare_lint! {
-    #[allow(text_direction_codepoint_in_literal)]
     /// The `text_direction_codepoint_in_comment` lint detects Unicode codepoints in comments that
     /// change the visual representation of text on screen in a way that does not correspond to
     /// their on memory representation.
@@ -3792,7 +3792,7 @@ declare_lint! {
     /// ```rust,compile_fail
     /// #![deny(text_direction_codepoint_in_comment)]
     /// fn main() {
-    ///     println!("{:?}"); // '‮');
+    #[doc = "    println!(\"{:?}\"); // '\u{202E}');"]
     /// }
     /// ```
     ///
@@ -3807,7 +3807,43 @@ declare_lint! {
     /// their use.
     pub TEXT_DIRECTION_CODEPOINT_IN_COMMENT,
     Deny,
-    "invisible directionality-changing codepoints in comment"
+    "invisible directionality-changing codepoints in comment",
+    crate_level_only
+}
+
+declare_lint! {
+    /// The `text_direction_codepoint_in_literal` lint detects Unicode codepoints that change the
+    /// visual representation of text on screen in a way that does not correspond to their on
+    /// memory representation.
+    ///
+    /// ### Explanation
+    ///
+    /// The unicode characters `\u{202A}`, `\u{202B}`, `\u{202D}`, `\u{202E}`, `\u{2066}`,
+    /// `\u{2067}`, `\u{2068}`, `\u{202C}` and `\u{2069}` make the flow of text on screen change
+    /// its direction on software that supports these codepoints. This makes the text "abc" display
+    /// as "cba" on screen. By leveraging software that supports these, people can write specially
+    /// crafted literals that make the surrounding code seem like it's performing one action, when
+    /// in reality it is performing another. Because of this, we proactively lint against their
+    /// presence to avoid surprises.
+    ///
+    /// ### Example
+    ///
+    /// ```rust,compile_fail
+    /// #![deny(text_direction_codepoint_in_literal)]
+    /// fn main() {
+    // ` - convince tidy that backticks match
+    #[doc = "    println!(\"{:?}\", '\u{202E}');"]
+    // `
+    /// }
+    /// ```
+    ///
+    /// {{produces}}
+    ///
+    pub TEXT_DIRECTION_CODEPOINT_IN_LITERAL,
+    Deny,
+    "detect special Unicode codepoints that affect the visual representation of text on screen, \
+     changing the direction in which text flows",
+    crate_level_only
 }
 
 declare_lint! {

compiler/rustc_lint_defs/src/builtin.rs

@@ -698,6 +698,14 @@ pub enum BuiltinLintDiag {
         is_string: bool,
         suggestion: Span,
     },
+    HiddenUnicodeCodepoints {
+        label: String,
+        count: usize,
+        span_label: Span,
+        labels: Option<Vec<(char, Span)>>,
+        escape: bool,
+        spans: Vec<(char, Span)>,
+    },
     TrailingMacro(bool, Ident),
     BreakWithLabelAndLoop(Span),
     UnicodeTextFlow(Span, String),