Auto merge of rust-lang#114569 - matthiaskrgr:rollup-p8tcxtz, r=matthiaskrgr

Rollup of 8 pull requests

Successful merges:

 - rust-lang#98935 (Implement `Option::take_if`)
 - rust-lang#114093 (Add regression test for `echo 'mod unknown;' | rustc -`)
 - rust-lang#114229 (Nest tests/codegen/sanitizer*.rs tests in sanitizer dir)
 - rust-lang#114230 (Nest other codegen test topics)
 - rust-lang#114362 (string.rs: remove "Basic usage" text)
 - rust-lang#114365 (str.rs: remove "Basic usage" text)
 - rust-lang#114382 (Add a new `compare_bytes` intrinsic instead of calling `memcmp` directly)
 - rust-lang#114549 (Style fix and refactor on resolve diagnostics)

r? `@ghost`
`@rustbot` modify labels: rollup

Author: bors

compiler/rustc_codegen_cranelift/src/intrinsics/mod.rs

@@ -1155,6 +1155,20 @@ fn codegen_regular_intrinsic_call<'tcx>(
             ret.write_cvalue(fx, CValue::by_val(is_eq_value, ret.layout()));
         }
 
+        sym::compare_bytes => {
+            intrinsic_args!(fx, args => (lhs_ptr, rhs_ptr, bytes_val); intrinsic);
+            let lhs_ptr = lhs_ptr.load_scalar(fx);
+            let rhs_ptr = rhs_ptr.load_scalar(fx);
+            let bytes_val = bytes_val.load_scalar(fx);
+
+            let params = vec![AbiParam::new(fx.pointer_type); 3];
+            let returns = vec![AbiParam::new(types::I32)];
+            let args = &[lhs_ptr, rhs_ptr, bytes_val];
+            // Here we assume that the `memcmp` provided by the target is a NOP for size 0.
+            let cmp = fx.lib_call("memcmp", params, returns, args)[0];
+            ret.write_cvalue(fx, CValue::by_val(cmp, ret.layout()));
+        }
+
         sym::const_allocate => {
             intrinsic_args!(fx, args => (_size, _align); intrinsic);
 

compiler/rustc_codegen_gcc/src/intrinsic/mod.rs

@@ -302,6 +302,21 @@ impl<'a, 'gcc, 'tcx> IntrinsicCallMethods<'tcx> for Builder<'a, 'gcc, 'tcx> {
                     }
                 }
 
+                sym::compare_bytes => {
+                    let a = args[0].immediate();
+                    let b = args[1].immediate();
+                    let n = args[2].immediate();
+
+                    let void_ptr_type = self.context.new_type::<*const ()>();
+                    let a_ptr = self.bitcast(a, void_ptr_type);
+                    let b_ptr = self.bitcast(b, void_ptr_type);
+
+                    // Here we assume that the `memcmp` provided by the target is a NOP for size 0.
+                    let builtin = self.context.get_builtin_function("memcmp");
+                    let cmp = self.context.new_call(None, builtin, &[a_ptr, b_ptr, n]);
+                    self.sext(cmp, self.type_ix(32))
+                }
+
                 sym::black_box => {
                     args[0].val.store(self, result);
 

compiler/rustc_codegen_llvm/src/context.rs

@@ -891,7 +891,8 @@ impl<'ll> CodegenCx<'ll, '_> {
         ifn!("llvm.prefetch", fn(ptr, t_i32, t_i32, t_i32) -> void);
 
         // This isn't an "LLVM intrinsic", but LLVM's optimization passes
-        // recognize it like one and we assume it exists in `core::slice::cmp`
+        // recognize it like one (including turning it into `bcmp` sometimes)
+        // and we use it to implement intrinsics like `raw_eq` and `compare_bytes`
         match self.sess().target.arch.as_ref() {
             "avr" | "msp430" => ifn!("memcmp", fn(ptr, ptr, t_isize) -> t_i16),
             _ => ifn!("memcmp", fn(ptr, ptr, t_isize) -> t_i32),

compiler/rustc_codegen_llvm/src/intrinsic.rs

@@ -329,6 +329,16 @@ impl<'ll, 'tcx> IntrinsicCallMethods<'tcx> for Builder<'_, 'll, 'tcx> {
                 }
             }
 
+            sym::compare_bytes => {
+                // Here we assume that the `memcmp` provided by the target is a NOP for size 0.
+                let cmp = self.call_intrinsic(
+                    "memcmp",
+                    &[args[0].immediate(), args[1].immediate(), args[2].immediate()],
+                );
+                // Some targets have `memcmp` returning `i16`, but the intrinsic is always `i32`.
+                self.sext(cmp, self.type_ix(32))
+            }
+
             sym::black_box => {
                 args[0].val.store(self, result);
                 let result_val_span = [result.llval];

compiler/rustc_const_eval/src/interpret/intrinsics.rs

@@ -261,6 +261,10 @@ impl<'mir, 'tcx: 'mir, M: Machine<'mir, 'tcx>> InterpCx<'mir, 'tcx, M> {
             sym::write_bytes => {
                 self.write_bytes_intrinsic(&args[0], &args[1], &args[2])?;
             }
+            sym::compare_bytes => {
+                let result = self.compare_bytes_intrinsic(&args[0], &args[1], &args[2])?;
+                self.write_scalar(result, dest)?;
+            }
             sym::arith_offset => {
                 let ptr = self.read_pointer(&args[0])?;
                 let offset_count = self.read_target_isize(&args[1])?;
@@ -643,6 +647,24 @@ impl<'mir, 'tcx: 'mir, M: Machine<'mir, 'tcx>> InterpCx<'mir, 'tcx, M> {
         self.write_bytes_ptr(dst, bytes)
     }
 
+    pub(crate) fn compare_bytes_intrinsic(
+        &mut self,
+        left: &OpTy<'tcx, <M as Machine<'mir, 'tcx>>::Provenance>,
+        right: &OpTy<'tcx, <M as Machine<'mir, 'tcx>>::Provenance>,
+        byte_count: &OpTy<'tcx, <M as Machine<'mir, 'tcx>>::Provenance>,
+    ) -> InterpResult<'tcx, Scalar<M::Provenance>> {
+        let left = self.read_pointer(left)?;
+        let right = self.read_pointer(right)?;
+        let n = Size::from_bytes(self.read_target_usize(byte_count)?);
+
+        let left_bytes = self.read_bytes_ptr_strip_provenance(left, n)?;
+        let right_bytes = self.read_bytes_ptr_strip_provenance(right, n)?;
+
+        // `Ordering`'s discriminants are -1/0/+1, so casting does the right thing.
+        let result = Ord::cmp(left_bytes, right_bytes) as i32;
+        Ok(Scalar::from_i32(result))
+    }
+
     pub(crate) fn raw_eq_intrinsic(
         &mut self,
         lhs: &OpTy<'tcx, <M as Machine<'mir, 'tcx>>::Provenance>,

compiler/rustc_hir_analysis/src/check/intrinsic.rs

@@ -273,6 +273,10 @@ pub fn check_intrinsic_type(tcx: TyCtxt<'_>, it: &hir::ForeignItem<'_>) {
                 ],
                 Ty::new_unit(tcx),
             ),
+            sym::compare_bytes => {
+                let byte_ptr = Ty::new_imm_ptr(tcx, tcx.types.u8);
+                (0, vec![byte_ptr, byte_ptr, tcx.types.usize], tcx.types.i32)
+            }
             sym::write_bytes | sym::volatile_set_memory => (
                 1,
                 vec![

compiler/rustc_resolve/src/late/diagnostics.rs

@@ -555,7 +555,6 @@ impl<'a: 'ast, 'ast, 'tcx> LateResolutionVisitor<'a, '_, 'ast, 'tcx> {
                 }
             })
             .collect::<Vec<_>>();
-        let crate_def_id = CRATE_DEF_ID.to_def_id();
         // Try to filter out intrinsics candidates, as long as we have
         // some other candidates to suggest.
         let intrinsic_candidates: Vec<_> = candidates
@@ -566,8 +565,9 @@ impl<'a: 'ast, 'ast, 'tcx> LateResolutionVisitor<'a, '_, 'ast, 'tcx> {
             .collect();
         if candidates.is_empty() {
             // Put them back if we have no more candidates to suggest...
-            candidates.extend(intrinsic_candidates);
+            candidates = intrinsic_candidates;
         }
+        let crate_def_id = CRATE_DEF_ID.to_def_id();
         if candidates.is_empty() && is_expected(Res::Def(DefKind::Enum, crate_def_id)) {
             let mut enum_candidates: Vec<_> = self
                 .r
@@ -1180,37 +1180,34 @@ impl<'a: 'ast, 'ast, 'tcx> LateResolutionVisitor<'a, '_, 'ast, 'tcx> {
     /// return the span of whole call and the span for all arguments expect the first one (`self`).
     fn call_has_self_arg(&self, source: PathSource<'_>) -> Option<(Span, Option<Span>)> {
         let mut has_self_arg = None;
-        if let PathSource::Expr(Some(parent)) = source {
-            match &parent.kind {
-                ExprKind::Call(_, args) if !args.is_empty() => {
-                    let mut expr_kind = &args[0].kind;
-                    loop {
-                        match expr_kind {
-                            ExprKind::Path(_, arg_name) if arg_name.segments.len() == 1 => {
-                                if arg_name.segments[0].ident.name == kw::SelfLower {
-                                    let call_span = parent.span;
-                                    let tail_args_span = if args.len() > 1 {
-                                        Some(Span::new(
-                                            args[1].span.lo(),
-                                            args.last().unwrap().span.hi(),
-                                            call_span.ctxt(),
-                                            None,
-                                        ))
-                                    } else {
-                                        None
-                                    };
-                                    has_self_arg = Some((call_span, tail_args_span));
-                                }
-                                break;
+        if let PathSource::Expr(Some(parent)) = source
+            && let ExprKind::Call(_, args) = &parent.kind
+            && !args.is_empty() {
+                let mut expr_kind = &args[0].kind;
+                loop {
+                    match expr_kind {
+                        ExprKind::Path(_, arg_name) if arg_name.segments.len() == 1 => {
+                            if arg_name.segments[0].ident.name == kw::SelfLower {
+                                let call_span = parent.span;
+                                let tail_args_span = if args.len() > 1 {
+                                    Some(Span::new(
+                                        args[1].span.lo(),
+                                        args.last().unwrap().span.hi(),
+                                        call_span.ctxt(),
+                                        None,
+                                    ))
+                                } else {
+                                    None
+                                };
+                                has_self_arg = Some((call_span, tail_args_span));
                             }
-                            ExprKind::AddrOf(_, _, expr) => expr_kind = &expr.kind,
-                            _ => break,
+                            break;
                         }
+                        ExprKind::AddrOf(_, _, expr) => expr_kind = &expr.kind,
+                        _ => break,
                     }
                 }
-                _ => (),
-            }
-        };
+        }
         has_self_arg
     }
 
@@ -1220,15 +1217,15 @@ impl<'a: 'ast, 'ast, 'tcx> LateResolutionVisitor<'a, '_, 'ast, 'tcx> {
         // where a brace being opened means a block is being started. Look
         // ahead for the next text to see if `span` is followed by a `{`.
         let sm = self.r.tcx.sess.source_map();
-        let sp = sm.span_look_ahead(span, None, Some(50));
-        let followed_by_brace = matches!(sm.span_to_snippet(sp), Ok(ref snippet) if snippet == "{");
-        // In case this could be a struct literal that needs to be surrounded
-        // by parentheses, find the appropriate span.
-        let closing_span = sm.span_look_ahead(span, Some("}"), Some(50));
-        let closing_brace: Option<Span> = sm
-            .span_to_snippet(closing_span)
-            .map_or(None, |s| if s == "}" { Some(span.to(closing_span)) } else { None });
-        (followed_by_brace, closing_brace)
+        if let Some(followed_brace_span) = sm.span_look_ahead(span, "{", Some(50)) {
+            // In case this could be a struct literal that needs to be surrounded
+            // by parentheses, find the appropriate span.
+            let close_brace_span = sm.span_look_ahead(followed_brace_span, "}", Some(50));
+            let closing_brace = close_brace_span.map(|sp| span.to(sp));
+            (true, closing_brace)
+        } else {
+            (false, None)
+        }
     }
 
     /// Provides context-dependent help for errors reported by the `smart_resolve_path_fragment`

compiler/rustc_span/src/source_map.rs

@@ -973,24 +973,21 @@ impl SourceMap {
         Span::new(BytePos(start_of_next_point), end_of_next_point, sp.ctxt(), None)
     }
 
-    /// Returns a new span to check next none-whitespace character or some specified expected character
-    /// If `expect` is none, the first span of non-whitespace character is returned.
-    /// If `expect` presented, the first span of the character `expect` is returned
-    /// Otherwise, the span reached to limit is returned.
-    pub fn span_look_ahead(&self, span: Span, expect: Option<&str>, limit: Option<usize>) -> Span {
+    /// Check whether span is followed by some specified expected string in limit scope
+    pub fn span_look_ahead(&self, span: Span, expect: &str, limit: Option<usize>) -> Option<Span> {
         let mut sp = span;
         for _ in 0..limit.unwrap_or(100_usize) {
             sp = self.next_point(sp);
             if let Ok(ref snippet) = self.span_to_snippet(sp) {
-                if expect.is_some_and(|es| snippet == es) {
-                    break;
+                if snippet == expect {
+                    return Some(sp);
                 }
-                if expect.is_none() && snippet.chars().any(|c| !c.is_whitespace()) {
+                if snippet.chars().any(|c| !c.is_whitespace()) {
                     break;
                 }
             }
         }
-        sp
+        None
     }
 
     /// Finds the width of the character, either before or after the end of provided span,

compiler/rustc_span/src/symbol.rs

@@ -498,6 +498,7 @@ symbols! {
         cold,
         collapse_debuginfo,
         column,
+        compare_bytes,
         compare_exchange,
         compare_exchange_weak,
         compile_error,

compiler/rustc_trait_selection/src/traits/error_reporting/suggestions.rs

@@ -1429,20 +1429,18 @@ impl<'tcx> TypeErrCtxtExt<'tcx> for TypeErrCtxt<'_, 'tcx> {
 
                         // Issue #109436, we need to add parentheses properly for method calls
                         // for example, `foo.into()` should be `(&foo).into()`
-                        if let Ok(snippet) = self.tcx.sess.source_map().span_to_snippet(
-                            self.tcx.sess.source_map().span_look_ahead(span, Some("."), Some(50)),
-                        ) {
-                            if snippet == "." {
-                                err.multipart_suggestion_verbose(
-                                    sugg_msg,
-                                    vec![
-                                        (span.shrink_to_lo(), format!("({sugg_prefix}")),
-                                        (span.shrink_to_hi(), ")".to_string()),
-                                    ],
-                                    Applicability::MaybeIncorrect,
-                                );
-                                return true;
-                            }
+                        if let Some(_) =
+                            self.tcx.sess.source_map().span_look_ahead(span, ".", Some(50))
+                        {
+                            err.multipart_suggestion_verbose(
+                                sugg_msg,
+                                vec![
+                                    (span.shrink_to_lo(), format!("({sugg_prefix}")),
+                                    (span.shrink_to_hi(), ")".to_string()),
+                                ],
+                                Applicability::MaybeIncorrect,
+                            );
+                            return true;
                         }
 
                         // Issue #104961, we need to add parentheses properly for compound expressions