-Zunsigned-char=unsigned|signed|default flag for c_char->u8/i8 selection override

Author: azhogin

compiler/rustc_session/src/config.rs

@@ -23,7 +23,7 @@ use rustc_macros::{Decodable, Encodable, HashStable_Generic};
 use rustc_span::edition::{DEFAULT_EDITION, EDITION_NAME_LIST, Edition, LATEST_STABLE_EDITION};
 use rustc_span::source_map::FilePathMapping;
 use rustc_span::{
-    FileName, FileNameDisplayPreference, RealFileName, SourceFileHashAlgorithm, Symbol, sym,
+    FileName, FileNameDisplayPreference, RealFileName, SourceFileHashAlgorithm, Symbol, kw, sym,
 };
 use rustc_target::spec::{
     FramePointer, LinkSelfContainedComponents, LinkerFeatures, SplitDebuginfo, Target, TargetTuple,
@@ -2946,7 +2946,7 @@ pub(crate) mod dep_tracking {
         LtoCli, MirStripDebugInfo, NextSolverConfig, OomStrategy, OptLevel, OutFileName,
         OutputType, OutputTypes, PatchableFunctionEntry, Polonius, RemapPathScopeComponents,
         ResolveDocLinks, SourceFileHashAlgorithm, SplitDwarfKind, SwitchWithOptPath,
-        SymbolManglingVersion, WasiExecModel,
+        SymbolManglingVersion, UnsignedCharVar, WasiExecModel,
     };
     use crate::lint;
     use crate::utils::NativeLib;
@@ -3049,6 +3049,7 @@ pub(crate) mod dep_tracking {
         FunctionReturn,
         WasmCAbi,
         Align,
+        UnsignedCharVar,
     );
 
     impl<T1, T2> DepTrackingHash for (T1, T2)
@@ -3323,3 +3324,31 @@ impl MirIncludeSpans {
         self == MirIncludeSpans::On
     }
 }
+
+/// The different settings that the `-Zunsigned-char` flag can have.
+#[derive(Clone, Copy, PartialEq, Hash, Debug, Default)]
+pub enum UnsignedCharVar {
+    /// Use default signed/unsigned c_char according to target configuration
+    #[default]
+    Default,
+
+    /// Set c_char to signed i8
+    Signed,
+
+    /// Set c_char to unsigned u8
+    Unsigned,
+}
+
+impl UnsignedCharVar {
+    pub const fn desc_symbol(&self) -> Symbol {
+        match *self {
+            Self::Default => kw::Default,
+            Self::Signed => sym::signed,
+            Self::Unsigned => sym::unsigned,
+        }
+    }
+
+    pub const fn all() -> [Symbol; 3] {
+        [Self::Unsigned.desc_symbol(), Self::Signed.desc_symbol(), Self::Default.desc_symbol()]
+    }
+}

compiler/rustc_session/src/config/cfg.rs

@@ -32,7 +32,7 @@ use rustc_span::{Symbol, sym};
 use rustc_target::spec::{PanicStrategy, RelocModel, SanitizerSet, Target};
 
 use crate::Session;
-use crate::config::{CrateType, FmtDebug};
+use crate::config::{CrateType, FmtDebug, UnsignedCharVar};
 
 /// The parsed `--cfg` options that define the compilation environment of the
 /// crate, used to drive conditional compilation.
@@ -144,6 +144,7 @@ pub(crate) fn disallow_cfgs(sess: &Session, user_cfgs: &Cfg) {
             | (sym::target_has_atomic_load_store, Some(_))
             | (sym::target_thread_local, None) => disallow(cfg, "--target"),
             (sym::fmt_debug, None | Some(_)) => disallow(cfg, "-Z fmt-debug"),
+            (sym::unsigned_char, None | Some(_)) => disallow(cfg, "-Z unsigned-char"),
             (sym::emscripten_wasm_eh, None | Some(_)) => disallow(cfg, "-Z emscripten_wasm_eh"),
             _ => {}
         }
@@ -306,6 +307,8 @@ pub(crate) fn default_configuration(sess: &Session) -> Cfg {
         ins_none!(sym::contract_checks);
     }
 
+    ins_sym!(sym::unsigned_char, sess.opts.unstable_opts.unsigned_char.desc_symbol());
+
     ret
 }
 
@@ -470,5 +473,7 @@ impl CheckCfg {
 
         ins!(sym::unix, no_values);
         ins!(sym::windows, no_values);
+
+        ins!(sym::unsigned_char, empty_values).extend(UnsignedCharVar::all());
     }
 }

compiler/rustc_session/src/options.rs

@@ -794,6 +794,7 @@ mod desc {
     pub(crate) const parse_mir_include_spans: &str =
         "either a boolean (`yes`, `no`, `on`, `off`, etc), or `nll` (default: `nll`)";
     pub(crate) const parse_align: &str = "a number that is a power of 2 between 1 and 2^29";
+    pub(crate) const parse_unsigned_char_var: &str = "one of `default`, `unsigned`, `signed`";
 }
 
 pub mod parse {
@@ -954,6 +955,16 @@ pub mod parse {
         true
     }
 
+    pub(crate) fn parse_unsigned_char_var(slot: &mut UnsignedCharVar, v: Option<&str>) -> bool {
+        *slot = match v {
+            Some("unsigned") => UnsignedCharVar::Unsigned,
+            Some("signed") => UnsignedCharVar::Signed,
+            Some("default") => UnsignedCharVar::Default,
+            _ => return false,
+        };
+        true
+    }
+
     pub(crate) fn parse_location_detail(ld: &mut LocationDetail, v: Option<&str>) -> bool {
         if let Some(v) = v {
             ld.line = false;
@@ -2567,6 +2578,8 @@ written to standard error output)"),
         `hir-tree` (dump the raw HIR),
         `thir-tree`, `thir-flat`,
         `mir` (the MIR), or `mir-cfg` (graphviz formatted MIR)"),
+    unsigned_char: UnsignedCharVar = (UnsignedCharVar::default(), parse_unsigned_char_var, [TRACKED TARGET_MODIFIER],
+        "Make c_char type unsigned [default|signed|unsigned]"),
     unsound_mir_opts: bool = (false, parse_bool, [TRACKED],
         "enable unsound and buggy MIR optimizations (default: no)"),
     /// This name is kind of confusing: Most unstable options enable something themselves, while

compiler/rustc_span/src/symbol.rs

@@ -1847,6 +1847,7 @@ symbols! {
         shr_assign,
         sig_dfl,
         sig_ign,
+        signed,
         simd,
         simd_add,
         simd_and,
@@ -2160,6 +2161,8 @@ symbols! {
         unsafe_fields,
         unsafe_no_drop_flag,
         unsafe_pin_internals,
+        unsigned,
+        unsigned_char,
         unsize,
         unsized_const_param_ty,
         unsized_const_params,

library/core/Cargo.toml

@@ -36,4 +36,5 @@ check-cfg = [
     # and to stdarch `core_arch` crate which messes-up with Cargo list
     # of declared features, we therefor expect any feature cfg
     'cfg(feature, values(any()))',
+    'cfg(unsigned_char, values(any()))'
 ]

library/core/src/ffi/primitives.rs

@@ -102,7 +102,8 @@ mod c_char_definition {
         //   architecture defaults). As we only have a target for userspace apps so there are no
         //   special cases for L4Re below.
         //   https://github.com/rust-lang/rust/pull/132975#issuecomment-2484645240
-        if #[cfg(all(
+        if #[cfg(any(unsigned_char = "unsigned", all(
+            not(unsigned_char = "signed"),
             not(windows),
             not(target_vendor = "apple"),
             any(
@@ -118,7 +119,7 @@ mod c_char_definition {
                 target_arch = "s390x",
                 target_arch = "xtensa",
             )
-        ))] {
+        )))] {
             pub(super) type c_char = u8;
         } else {
             // On every other target, c_char is signed.

tests/ui/cfg/disallowed-cli-cfgs.rs

@@ -8,6 +8,7 @@
 //@ revisions: target_thread_local_ relocation_model_
 //@ revisions: fmt_debug_
 //@ revisions: emscripten_wasm_eh_
+//@ revisions: unsigned_char_
 
 //@ [overflow_checks_]compile-flags: --cfg overflow_checks
 //@ [debug_assertions_]compile-flags: --cfg debug_assertions
@@ -35,5 +36,6 @@
 //@ [relocation_model_]compile-flags: --cfg relocation_model="a"
 //@ [fmt_debug_]compile-flags: --cfg fmt_debug="shallow"
 //@ [emscripten_wasm_eh_]compile-flags: --cfg emscripten_wasm_eh
+//@ [unsigned_char_]compile-flags: --cfg unsigned_char="unsigned"
 
 fn main() {}

tests/ui/cfg/disallowed-cli-cfgs.unsigned_char_.stderr

@@ -0,0 +1,8 @@
+error: unexpected `--cfg unsigned_char="unsigned"` flag
+   |
+   = note: config `unsigned_char` is only supposed to be controlled by `-Z unsigned-char`
+   = note: manually setting a built-in cfg can and does create incoherent behaviors
+   = note: `#[deny(explicit_builtin_cfgs_in_flags)]` on by default
+
+error: aborting due to 1 previous error
+

tests/ui/check-cfg/well-known-names.stderr

@@ -33,7 +33,8 @@ LL | #[cfg(list_all_well_known_cfgs)]
 `target_thread_local`
 `target_vendor`
 `ub_checks`
-`unix`, and `windows`
+`unix`
+`unsigned_char`, and `windows`
    = help: to expect this configuration use `--check-cfg=cfg(list_all_well_known_cfgs)`
    = note: see <https://doc.rust-lang.org/nightly/rustc/check-cfg.html> for more information about checking conditional configuration
    = note: `#[warn(unexpected_cfgs)]` on by default

tests/ui/check-cfg/well-known-values.rs

@@ -80,6 +80,8 @@
     //~^ WARN unexpected `cfg` condition value
     unix = "_UNEXPECTED_VALUE",
     //~^ WARN unexpected `cfg` condition value
+    unsigned_char = "_UNEXPECTED_VALUE",
+    //~^ WARN unexpected `cfg` condition value
     windows = "_UNEXPECTED_VALUE",
     //~^ WARN unexpected `cfg` condition value
     // tidy-alphabetical-end

tests/ui/check-cfg/well-known-values.stderr

@@ -258,6 +258,15 @@ LL |     unix = "_UNEXPECTED_VALUE",
 warning: unexpected `cfg` condition value: `_UNEXPECTED_VALUE`
   --> $DIR/well-known-values.rs:83:5
    |
+LL |     unsigned_char = "_UNEXPECTED_VALUE",
+   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+   |
+   = note: expected values for `unsigned_char` are: `default`, `signed`, and `unsigned`
+   = note: see <https://doc.rust-lang.org/nightly/rustc/check-cfg.html> for more information about checking conditional configuration
+
+warning: unexpected `cfg` condition value: `_UNEXPECTED_VALUE`
+  --> $DIR/well-known-values.rs:85:5
+   |
 LL |     windows = "_UNEXPECTED_VALUE",
    |     ^^^^^^^----------------------
    |            |
@@ -267,7 +276,7 @@ LL |     windows = "_UNEXPECTED_VALUE",
    = note: see <https://doc.rust-lang.org/nightly/rustc/check-cfg.html> for more information about checking conditional configuration
 
 warning: unexpected `cfg` condition value: `linuz`
-  --> $DIR/well-known-values.rs:89:7
+  --> $DIR/well-known-values.rs:91:7
    |
 LL | #[cfg(target_os = "linuz")] // testing that we suggest `linux`
    |       ^^^^^^^^^^^^-------
@@ -277,5 +286,5 @@ LL | #[cfg(target_os = "linuz")] // testing that we suggest `linux`
    = note: expected values for `target_os` are: `aix`, `amdhsa`, `android`, `cuda`, `cygwin`, `dragonfly`, `emscripten`, `espidf`, `freebsd`, `fuchsia`, `haiku`, `hermit`, `horizon`, `hurd`, `illumos`, `ios`, `l4re`, `linux`, `macos`, `netbsd`, `none`, `nto`, `nuttx`, `openbsd`, `psp`, `psx`, `redox`, `rtems`, `solaris`, `solid_asp3`, `teeos`, `trusty`, `tvos`, `uefi`, `unknown`, `visionos`, `vita`, `vxworks`, `wasi`, `watchos`, `windows`, `xous`, and `zkvm`
    = note: see <https://doc.rust-lang.org/nightly/rustc/check-cfg.html> for more information about checking conditional configuration
 
-warning: 28 warnings emitted
+warning: 29 warnings emitted
 

tests/ui/precondition-checks/cfg-unsigned-char.rs

@@ -0,0 +1,48 @@
+//@ build-pass
+//@ revisions: unsigned_ signed_ default_
+//@ [unsigned_] compile-flags: -Zunsigned-char=unsigned
+//@ [signed_] compile-flags: -Zunsigned-char=signed
+//@ [default_] compile-flags: -Zunsigned-char=default
+
+#![no_core]
+#![crate_type = "rlib"]
+#![feature(intrinsics, rustc_attrs, no_core, lang_items, staged_api)]
+#![stable(feature = "test", since = "1.0.0")]
+
+#[lang="sized"]
+trait Sized {}
+#[lang = "copy"]
+trait Copy {}
+impl Copy for bool {}
+
+#[stable(feature = "test", since = "1.0.0")]
+#[rustc_const_stable(feature = "test", since = "1.0.0")]
+#[rustc_intrinsic]
+const unsafe fn unreachable() -> !;
+
+#[rustc_builtin_macro]
+macro_rules! cfg {
+    ($($cfg:tt)*) => {};
+}
+
+const fn do_or_die(cond: bool) {
+    if cond {
+    } else {
+        unsafe { unreachable() }
+    }
+}
+
+macro_rules! assert {
+    ($x:expr $(,)?) => {
+        const _: () = do_or_die($x);
+    };
+}
+
+fn main() {
+    #[cfg(unsigned_)]
+    assert!(cfg!(unsigned_char = "unsigned"));
+    #[cfg(signed_)]
+    assert!(cfg!(unsigned_char = "signed"));
+    #[cfg(default_)]
+    assert!(cfg!(unsigned_char = "default"));
+}