Auto merge of #87194 - eddyb:const-value-mangling, r=michaelwoerister,oli-obk

rustc_symbol_mangling: support structural constants and &str in v0.

This PR should unblock #85530 (except for float `const` generics, which AFAIK should've never worked).
(cc `@tmiasko` could the #85530 (comment) failures be retried with a quick crater "subset" run of this PR + changing the default to `v0`? Just to make sure I didn't miss anything other than the floats)

The encoding is the one suggested before in e.g. #61486 (comment), tho this PR won't by itself finish #61486, before closing that we'd likely want to move to `@oli-obk's` "valtrees" (i.e. #83234 and other associated work).

<hr>

**EDITs**:
1. switched unit/tuple/braced-with-named-fields `<const-fields>` prefixes from `"u"`/`"T"`/`""` to `"U"`/`"T"`/`"S"` to avoid the ambiguity reported by `@tmiasko` in #87194 (comment).

2. `rustc-demangle` PR: rust-lang/rustc-demangle#55

3. RFC amendment PR: rust-lang/rfcs#3161
    * also removed the grammar changes included in that PR, from this description

4. added tests (temporarily using my fork of `rustc-demangle`)

<hr>

r? `@michaelwoerister`

Author: bors

Cargo.lock

@@ -3398,9 +3398,9 @@ dependencies = [
 
 [[package]]
 name = "rustc-demangle"
-version = "0.1.18"
+version = "0.1.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6e3bad0ee36814ca07d7968269dd4b7ec89ec2da10c4bb613928d3077083c232"
+checksum = "7ef03e0a2b150c7a90d01faf6254c9c48a41e95fb2a8c2ac1c6f0d2b9aefc342"
 dependencies = [
  "compiler_builtins",
  "rustc-std-workspace-core",

compiler/rustc_codegen_llvm/Cargo.toml

@@ -15,7 +15,7 @@ measureme = "9.1.0"
 snap = "1"
 tracing = "0.1"
 rustc_middle = { path = "../rustc_middle" }
-rustc-demangle = "0.1.18"
+rustc-demangle = "0.1.21"
 rustc_attr = { path = "../rustc_attr" }
 rustc_codegen_ssa = { path = "../rustc_codegen_ssa" }
 rustc_data_structures = { path = "../rustc_data_structures" }

compiler/rustc_middle/src/ty/print/pretty.rs

@@ -1220,13 +1220,20 @@ pub trait PrettyPrinter<'tcx>:
                         }
                         p!(")");
                     }
-                    ty::Adt(def, substs) if def.variants.is_empty() => {
-                        p!(print_value_path(def.did, substs));
+                    ty::Adt(def, _) if def.variants.is_empty() => {
+                        self = self.typed_value(
+                            |mut this| {
+                                write!(this, "unreachable()")?;
+                                Ok(this)
+                            },
+                            |this| this.print_type(ty),
+                            ": ",
+                        )?;
                     }
                     ty::Adt(def, substs) => {
-                        let variant_id =
-                            contents.variant.expect("destructed const of adt without variant id");
-                        let variant_def = &def.variants[variant_id];
+                        let variant_idx =
+                            contents.variant.expect("destructed const of adt without variant idx");
+                        let variant_def = &def.variants[variant_idx];
                         p!(print_value_path(variant_def.def_id, substs));
 
                         match variant_def.ctor_kind {

compiler/rustc_symbol_mangling/Cargo.toml

@@ -9,7 +9,7 @@ doctest = false
 [dependencies]
 tracing = "0.1"
 punycode = "0.4.0"
-rustc-demangle = "0.1.18"
+rustc-demangle = "0.1.21"
 
 rustc_span = { path = "../rustc_span" }
 rustc_middle = { path = "../rustc_middle" }

compiler/rustc_symbol_mangling/src/lib.rs

@@ -91,6 +91,7 @@
 #![feature(never_type)]
 #![feature(nll)]
 #![feature(in_band_lifetimes)]
+#![feature(iter_zip)]
 #![recursion_limit = "256"]
 
 #[macro_use]

compiler/rustc_symbol_mangling/src/v0.rs

@@ -1,8 +1,10 @@
 use rustc_data_structures::base_n;
 use rustc_data_structures::fx::{FxHashMap, FxHashSet};
 use rustc_hir as hir;
+use rustc_hir::def::CtorKind;
 use rustc_hir::def_id::{CrateNum, DefId};
 use rustc_hir::definitions::{DefPathData, DisambiguatedDefPathData};
+use rustc_middle::mir::interpret::ConstValue;
 use rustc_middle::ty::layout::IntegerExt;
 use rustc_middle::ty::print::{Print, Printer};
 use rustc_middle::ty::subst::{GenericArg, GenericArgKind, Subst};
@@ -11,6 +13,7 @@ use rustc_target::abi::Integer;
 use rustc_target::spec::abi::Abi;
 
 use std::fmt::Write;
+use std::iter;
 use std::ops::Range;
 
 pub(super) fn mangle(
@@ -534,39 +537,153 @@ impl Printer<'tcx> for &mut SymbolMangler<'tcx> {
     }
 
     fn print_const(mut self, ct: &'tcx ty::Const<'tcx>) -> Result<Self::Const, Self::Error> {
+        // We only mangle a typed value if the const can be evaluated.
+        let ct = ct.eval(self.tcx, ty::ParamEnv::reveal_all());
+        match ct.val {
+            ty::ConstKind::Value(_) => {}
+
+            // Placeholders (should be demangled as `_`).
+            // NOTE(eddyb) despite `Unevaluated` having a `DefId` (and therefore
+            // a path), even for it we still need to encode a placeholder, as
+            // the path could refer back to e.g. an `impl` using the constant.
+            ty::ConstKind::Unevaluated(_)
+            | ty::ConstKind::Param(_)
+            | ty::ConstKind::Infer(_)
+            | ty::ConstKind::Bound(..)
+            | ty::ConstKind::Placeholder(_)
+            | ty::ConstKind::Error(_) => {
+                // Never cached (single-character).
+                self.push("p");
+                return Ok(self);
+            }
+        }
+
         if let Some(&i) = self.consts.get(&ct) {
             return self.print_backref(i);
         }
         let start = self.out.len();
 
-        let mut neg = false;
-        let val = match ct.ty.kind() {
-            ty::Uint(_) | ty::Bool | ty::Char => {
-                ct.try_eval_bits(self.tcx, ty::ParamEnv::reveal_all(), ct.ty)
-            }
-            ty::Int(ity) => {
-                ct.try_eval_bits(self.tcx, ty::ParamEnv::reveal_all(), ct.ty).and_then(|b| {
-                    let val = Integer::from_int_ty(&self.tcx, *ity).size().sign_extend(b) as i128;
+        match ct.ty.kind() {
+            ty::Uint(_) | ty::Int(_) | ty::Bool | ty::Char => {
+                self = ct.ty.print(self)?;
+
+                let mut bits = ct.eval_bits(self.tcx, ty::ParamEnv::reveal_all(), ct.ty);
+
+                // Negative integer values are mangled using `n` as a "sign prefix".
+                if let ty::Int(ity) = ct.ty.kind() {
+                    let val =
+                        Integer::from_int_ty(&self.tcx, *ity).size().sign_extend(bits) as i128;
                     if val < 0 {
-                        neg = true;
+                        self.push("n");
                     }
-                    Some(val.unsigned_abs())
-                })
+                    bits = val.unsigned_abs();
+                }
+
+                let _ = write!(self.out, "{:x}_", bits);
             }
+
+            // HACK(eddyb) because `ty::Const` only supports sized values (for now),
+            // we can't use `deref_const` + supporting `str`, we have to specially
+            // handle `&str` and include both `&` ("R") and `str` ("e") prefixes.
+            ty::Ref(_, ty, hir::Mutability::Not) if *ty == self.tcx.types.str_ => {
+                self.push("R");
+                match ct.val {
+                    ty::ConstKind::Value(ConstValue::Slice { data, start, end }) => {
+                        // NOTE(eddyb) the following comment was kept from `ty::print::pretty`:
+                        // The `inspect` here is okay since we checked the bounds, and there are no
+                        // relocations (we have an active `str` reference here). We don't use this
+                        // result to affect interpreter execution.
+                        let slice =
+                            data.inspect_with_uninit_and_ptr_outside_interpreter(start..end);
+                        let s = std::str::from_utf8(slice).expect("non utf8 str from miri");
+
+                        self.push("e");
+                        // FIXME(eddyb) use a specialized hex-encoding loop.
+                        for byte in s.bytes() {
+                            let _ = write!(self.out, "{:02x}", byte);
+                        }
+                        self.push("_");
+                    }
+
+                    _ => {
+                        bug!("symbol_names: unsupported `&str` constant: {:?}", ct);
+                    }
+                }
+            }
+
+            ty::Ref(_, _, mutbl) => {
+                self.push(match mutbl {
+                    hir::Mutability::Not => "R",
+                    hir::Mutability::Mut => "Q",
+                });
+                self = self.tcx.deref_const(ty::ParamEnv::reveal_all().and(ct)).print(self)?;
+            }
+
+            ty::Array(..) | ty::Tuple(..) | ty::Adt(..) => {
+                let contents = self.tcx.destructure_const(ty::ParamEnv::reveal_all().and(ct));
+                let fields = contents.fields.iter().copied();
+
+                let print_field_list = |mut this: Self| {
+                    for field in fields.clone() {
+                        this = field.print(this)?;
+                    }
+                    this.push("E");
+                    Ok(this)
+                };
+
+                match *ct.ty.kind() {
+                    ty::Array(..) => {
+                        self.push("A");
+                        self = print_field_list(self)?;
+                    }
+                    ty::Tuple(..) => {
+                        self.push("T");
+                        self = print_field_list(self)?;
+                    }
+                    ty::Adt(def, substs) => {
+                        let variant_idx =
+                            contents.variant.expect("destructed const of adt without variant idx");
+                        let variant_def = &def.variants[variant_idx];
+
+                        self.push("V");
+                        self = self.print_def_path(variant_def.def_id, substs)?;
+
+                        match variant_def.ctor_kind {
+                            CtorKind::Const => {
+                                self.push("U");
+                            }
+                            CtorKind::Fn => {
+                                self.push("T");
+                                self = print_field_list(self)?;
+                            }
+                            CtorKind::Fictive => {
+                                self.push("S");
+                                for (field_def, field) in iter::zip(&variant_def.fields, fields) {
+                                    // HACK(eddyb) this mimics `path_append`,
+                                    // instead of simply using `field_def.ident`,
+                                    // just to be able to handle disambiguators.
+                                    let disambiguated_field =
+                                        self.tcx.def_key(field_def.did).disambiguated_data;
+                                    let field_name =
+                                        disambiguated_field.data.get_opt_name().map(|s| s.as_str());
+                                    self.push_disambiguator(
+                                        disambiguated_field.disambiguator as u64,
+                                    );
+                                    self.push_ident(&field_name.as_ref().map_or("", |s| &s[..]));
+
+                                    self = field.print(self)?;
+                                }
+                                self.push("E");
+                            }
+                        }
+                    }
+                    _ => unreachable!(),
+                }
+            }
+
             _ => {
                 bug!("symbol_names: unsupported constant of type `{}` ({:?})", ct.ty, ct);
             }
-        };
-
-        if let Some(bits) = val {
-            // We only print the type if the const can be evaluated.
-            self = ct.ty.print(self)?;
-            let _ = write!(self.out, "{}{:x}_", if neg { "n" } else { "" }, bits);
-        } else {
-            // NOTE(eddyb) despite having the path, we need to
-            // encode a placeholder, as the path could refer
-            // back to e.g. an `impl` using the constant.
-            self.push("p");
         }
 
         // Only cache consts that do not refer to an enclosing

library/std/Cargo.toml

@@ -24,7 +24,7 @@ std_detect = { path = "../stdarch/crates/std_detect", default-features = false,
 
 # Dependencies of the `backtrace` crate
 addr2line = { version = "0.16.0", optional = true, default-features = false }
-rustc-demangle = { version = "0.1.18", features = ['rustc-dep-of-std'] }
+rustc-demangle = { version = "0.1.21", features = ['rustc-dep-of-std'] }
 miniz_oxide = { version = "0.4.0", optional = true, default-features = false }
 [dependencies.object]
 version = "0.26.1"

src/test/debuginfo/function-call.rs

@@ -1,5 +1,5 @@
 // This test does not passed with gdb < 8.0. See #53497.
-// min-gdb-version: 8.0
+// min-gdb-version: 10.1
 
 // compile-flags:-g
 

src/test/debuginfo/function-names.rs

@@ -1,5 +1,5 @@
 // Function names are formatted differently in old versions of GDB
-// min-gdb-version: 9.2
+// min-gdb-version: 10.1
 
 // compile-flags:-g
 

src/test/ui/panics/issue-47429-short-backtraces.run.stderr src/test/ui/panics/issue-47429-short-backtraces.legacy.run.stderr

@@ -1,4 +1,4 @@
-thread 'main' panicked at 'explicit panic', $DIR/issue-47429-short-backtraces.rs:16:5
+thread 'main' panicked at 'explicit panic', $DIR/issue-47429-short-backtraces.rs:21:5
 stack backtrace:
    0: std::panicking::begin_panic
    1: issue_47429_short_backtraces::main

src/test/ui/panics/issue-47429-short-backtraces.rs

@@ -12,6 +12,11 @@
 // ignore-emscripten no panic or subprocess support
 // ignore-sgx no subprocess support
 
+// NOTE(eddyb) output differs between symbol mangling schemes
+// revisions: legacy v0
+// [legacy] compile-flags: -Zsymbol-mangling-version=legacy
+//     [v0] compile-flags: -Zsymbol-mangling-version=v0
+
 fn main() {
     panic!()
 }

src/test/ui/panics/issue-47429-short-backtraces.v0.run.stderr

@@ -0,0 +1,5 @@
+thread 'main' panicked at 'explicit panic', $DIR/issue-47429-short-backtraces.rs:21:5
+stack backtrace:
+   0: std::panicking::begin_panic::<&str>
+   1: issue_47429_short_backtraces::main
+note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

src/test/ui/symbol-names/const-generics-demangling.rs

@@ -1,37 +1,37 @@
 // build-fail
-// compile-flags: -Z symbol-mangling-version=v0
+// compile-flags: -Z symbol-mangling-version=v0 --crate-name=c
 #![feature(rustc_attrs)]
 
 pub struct Unsigned<const F: u8>;
 
 #[rustc_symbol_name]
-//~^ ERROR symbol-name(_RMCsaP8qXevlYG3_25const_generics_demanglingINtB0_8UnsignedKhb_E)
-//~| ERROR demangling(<const_generics_demangling[7e153590edc26969]::Unsigned<11: u8>>)
-//~| ERROR demangling-alt(<const_generics_demangling::Unsigned<11>>)
+//~^ ERROR symbol-name(_RMCsno73SFvQKx_1cINtB0_8UnsignedKhb_E)
+//~| ERROR demangling(<c[464da6a86cb672f]::Unsigned<11u8>>)
+//~| ERROR demangling-alt(<c::Unsigned<11>>)
 impl Unsigned<11> {}
 
 pub struct Signed<const F: i16>;
 
 #[rustc_symbol_name]
-//~^ ERROR symbol-name(_RMs_CsaP8qXevlYG3_25const_generics_demanglingINtB2_6SignedKsn98_E)
-//~| ERROR demangling(<const_generics_demangling[7e153590edc26969]::Signed<-152: i16>>)
-//~| ERROR demangling-alt(<const_generics_demangling::Signed<-152>>)
+//~^ ERROR symbol-name(_RMs_Csno73SFvQKx_1cINtB2_6SignedKsn98_E)
+//~| ERROR demangling(<c[464da6a86cb672f]::Signed<-152i16>>)
+//~| ERROR demangling-alt(<c::Signed<-152>>)
 impl Signed<-152> {}
 
 pub struct Bool<const F: bool>;
 
 #[rustc_symbol_name]
-//~^ ERROR symbol-name(_RMs0_CsaP8qXevlYG3_25const_generics_demanglingINtB3_4BoolKb1_E)
-//~| ERROR demangling(<const_generics_demangling[7e153590edc26969]::Bool<true: bool>>)
-//~| ERROR demangling-alt(<const_generics_demangling::Bool<true>>)
+//~^ ERROR symbol-name(_RMs0_Csno73SFvQKx_1cINtB3_4BoolKb1_E)
+//~| ERROR demangling(<c[464da6a86cb672f]::Bool<true>>)
+//~| ERROR demangling-alt(<c::Bool<true>>)
 impl Bool<true> {}
 
 pub struct Char<const F: char>;
 
 #[rustc_symbol_name]
-//~^ ERROR symbol-name(_RMs1_CsaP8qXevlYG3_25const_generics_demanglingINtB3_4CharKc2202_E)
-//~| ERROR demangling(<const_generics_demangling[7e153590edc26969]::Char<'∂': char>>)
-//~| ERROR demangling-alt(<const_generics_demangling::Char<'∂'>>)
+//~^ ERROR symbol-name(_RMs1_Csno73SFvQKx_1cINtB3_4CharKc2202_E)
+//~| ERROR demangling(<c[464da6a86cb672f]::Char<'∂'>>)
+//~| ERROR demangling-alt(<c::Char<'∂'>>)
 impl Char<'∂'> {}
 
 fn main() {}