Auto merge of rust-lang#77381 - Dylan-DPC:rollup-0sr6p5p, r=Dylan-DPC

Rollup of 12 pull requests

Successful merges:

 - rust-lang#76909 (Add Iterator::advance_by and DoubleEndedIterator::advance_back_by)
 - rust-lang#77153 (Fix recursive nonterminal expansion during pretty-print/reparse check)
 - rust-lang#77202 (Defer Apple SDKROOT detection to link time.)
 - rust-lang#77303 (const evaluatable: improve `TooGeneric` handling)
 - rust-lang#77305 (move candidate_from_obligation_no_cache)
 - rust-lang#77315 (Rename AllocErr to AllocError)
 - rust-lang#77319 (Stable hashing: add comments and tests concerning platform-independence)
 - rust-lang#77324 (Don't fire `const_item_mutation` lint on writes through a pointer)
 - rust-lang#77343 (Validate `rustc_args_required_const`)
 - rust-lang#77349 (Update cargo)
 - rust-lang#77360 (References to ZSTs may be at arbitrary aligned addresses)
 - rust-lang#77371 (Remove trailing space in error message)

Failed merges:

r? `@ghost`

compiler/rustc_codegen_ssa/src/back/link.rs

@@ -1524,6 +1524,9 @@ fn linker_with_args<'a, B: ArchiveBuilder<'a>>(
     // NO-OPT-OUT, OBJECT-FILES-MAYBE, CUSTOMIZATION-POINT
     add_pre_link_args(cmd, sess, flavor);
 
+    // NO-OPT-OUT, OBJECT-FILES-NO
+    add_apple_sdk(cmd, sess, flavor);
+
     // NO-OPT-OUT
     add_link_script(cmd, sess, tmpdir, crate_type);
 
@@ -2083,3 +2086,86 @@ fn are_upstream_rust_objects_already_included(sess: &Session) -> bool {
         config::Lto::No | config::Lto::ThinLocal => false,
     }
 }
+
+fn add_apple_sdk(cmd: &mut dyn Linker, sess: &Session, flavor: LinkerFlavor) {
+    let arch = &sess.target.target.arch;
+    let os = &sess.target.target.target_os;
+    let llvm_target = &sess.target.target.llvm_target;
+    if sess.target.target.target_vendor != "apple"
+        || !matches!(os.as_str(), "ios" | "tvos")
+        || flavor != LinkerFlavor::Gcc
+    {
+        return;
+    }
+    let sdk_name = match (arch.as_str(), os.as_str()) {
+        ("aarch64", "tvos") => "appletvos",
+        ("x86_64", "tvos") => "appletvsimulator",
+        ("arm", "ios") => "iphoneos",
+        ("aarch64", "ios") => "iphoneos",
+        ("x86", "ios") => "iphonesimulator",
+        ("x86_64", "ios") if llvm_target.contains("macabi") => "macosx10.15",
+        ("x86_64", "ios") => "iphonesimulator",
+        _ => {
+            sess.err(&format!("unsupported arch `{}` for os `{}`", arch, os));
+            return;
+        }
+    };
+    let sdk_root = match get_apple_sdk_root(sdk_name) {
+        Ok(s) => s,
+        Err(e) => {
+            sess.err(&e);
+            return;
+        }
+    };
+    let arch_name = llvm_target.split('-').next().expect("LLVM target must have a hyphen");
+    cmd.args(&["-arch", arch_name, "-isysroot", &sdk_root, "-Wl,-syslibroot", &sdk_root]);
+}
+
+fn get_apple_sdk_root(sdk_name: &str) -> Result<String, String> {
+    // Following what clang does
+    // (https://github.com/llvm/llvm-project/blob/
+    // 296a80102a9b72c3eda80558fb78a3ed8849b341/clang/lib/Driver/ToolChains/Darwin.cpp#L1661-L1678)
+    // to allow the SDK path to be set. (For clang, xcrun sets
+    // SDKROOT; for rustc, the user or build system can set it, or we
+    // can fall back to checking for xcrun on PATH.)
+    if let Ok(sdkroot) = env::var("SDKROOT") {
+        let p = Path::new(&sdkroot);
+        match sdk_name {
+            // Ignore `SDKROOT` if it's clearly set for the wrong platform.
+            "appletvos"
+                if sdkroot.contains("TVSimulator.platform")
+                    || sdkroot.contains("MacOSX.platform") => {}
+            "appletvsimulator"
+                if sdkroot.contains("TVOS.platform") || sdkroot.contains("MacOSX.platform") => {}
+            "iphoneos"
+                if sdkroot.contains("iPhoneSimulator.platform")
+                    || sdkroot.contains("MacOSX.platform") => {}
+            "iphonesimulator"
+                if sdkroot.contains("iPhoneOS.platform") || sdkroot.contains("MacOSX.platform") => {
+            }
+            "macosx10.15"
+                if sdkroot.contains("iPhoneOS.platform")
+                    || sdkroot.contains("iPhoneSimulator.platform") => {}
+            // Ignore `SDKROOT` if it's not a valid path.
+            _ if !p.is_absolute() || p == Path::new("/") || !p.exists() => {}
+            _ => return Ok(sdkroot),
+        }
+    }
+    let res =
+        Command::new("xcrun").arg("--show-sdk-path").arg("-sdk").arg(sdk_name).output().and_then(
+            |output| {
+                if output.status.success() {
+                    Ok(String::from_utf8(output.stdout).unwrap())
+                } else {
+                    let error = String::from_utf8(output.stderr);
+                    let error = format!("process exit with error: {}", error.unwrap());
+                    Err(io::Error::new(io::ErrorKind::Other, &error[..]))
+                }
+            },
+        );
+
+    match res {
+        Ok(output) => Ok(output.trim().to_string()),
+        Err(e) => Err(format!("failed to get {} SDK path: {}", sdk_name, e)),
+    }
+}

compiler/rustc_data_structures/src/sip128.rs

@@ -125,15 +125,28 @@ impl SipHasher128 {
 
     // A specialized write function for values with size <= 8.
     //
-    // The hashing of multi-byte integers depends on endianness. E.g.:
+    // The input must be zero-extended to 64-bits by the caller. This extension
+    // isn't hashed, but the implementation requires it for correctness.
+    //
+    // This function, given the same integer size and value, has the same effect
+    // on both little- and big-endian hardware. It operates on values without
+    // depending on their sequence in memory, so is independent of endianness.
+    //
+    // However, we want SipHasher128 to be platform-dependent, in order to be
+    // consistent with the platform-dependent SipHasher in libstd. In other
+    // words, we want:
+    //
     // - little-endian: `write_u32(0xDDCCBBAA)` == `write([0xAA, 0xBB, 0xCC, 0xDD])`
     // - big-endian:    `write_u32(0xDDCCBBAA)` == `write([0xDD, 0xCC, 0xBB, 0xAA])`
     //
-    // This function does the right thing for little-endian hardware. On
-    // big-endian hardware `x` must be byte-swapped first to give the right
-    // behaviour. After any byte-swapping, the input must be zero-extended to
-    // 64-bits. The caller is responsible for the byte-swapping and
-    // zero-extension.
+    // Therefore, in order to produce endian-dependent results, SipHasher128's
+    // `write_xxx` Hasher trait methods byte-swap `x` prior to zero-extending.
+    //
+    // If clients of SipHasher128 itself want platform-independent results, they
+    // *also* must byte-swap integer inputs before invoking the `write_xxx`
+    // methods on big-endian hardware (that is, two byte-swaps must occur--one
+    // in the client, and one in SipHasher128). Additionally, they must extend
+    // `usize` and `isize` types to 64 bits on 32-bit systems.
     #[inline]
     fn short_write<T>(&mut self, _x: T, x: u64) {
         let size = mem::size_of::<T>();

compiler/rustc_data_structures/src/sip128/tests.rs

@@ -1,7 +1,6 @@
 use super::*;
 
 use std::hash::{Hash, Hasher};
-use std::{mem, slice};
 
 // Hash just the bytes of the slice, without length prefix
 struct Bytes<'a>(&'a [u8]);
@@ -399,20 +398,55 @@ fn test_hash_no_concat_alias() {
 }
 
 #[test]
-fn test_write_short_works() {
-    let test_usize = 0xd0c0b0a0usize;
+fn test_short_write_works() {
+    let test_u8 = 0xFF_u8;
+    let test_u16 = 0x1122_u16;
+    let test_u32 = 0x22334455_u32;
+    let test_u64 = 0x33445566_778899AA_u64;
+    let test_u128 = 0x11223344_55667788_99AABBCC_DDEEFF77_u128;
+    let test_usize = 0xD0C0B0A0_usize;
+
+    let test_i8 = -1_i8;
+    let test_i16 = -2_i16;
+    let test_i32 = -3_i32;
+    let test_i64 = -4_i64;
+    let test_i128 = -5_i128;
+    let test_isize = -6_isize;
+
     let mut h1 = SipHasher128::new_with_keys(0, 0);
-    h1.write_usize(test_usize);
     h1.write(b"bytes");
     h1.write(b"string");
-    h1.write_u8(0xFFu8);
-    h1.write_u8(0x01u8);
+    h1.write_u8(test_u8);
+    h1.write_u16(test_u16);
+    h1.write_u32(test_u32);
+    h1.write_u64(test_u64);
+    h1.write_u128(test_u128);
+    h1.write_usize(test_usize);
+    h1.write_i8(test_i8);
+    h1.write_i16(test_i16);
+    h1.write_i32(test_i32);
+    h1.write_i64(test_i64);
+    h1.write_i128(test_i128);
+    h1.write_isize(test_isize);
+
     let mut h2 = SipHasher128::new_with_keys(0, 0);
-    h2.write(unsafe {
-        slice::from_raw_parts(&test_usize as *const _ as *const u8, mem::size_of::<usize>())
-    });
     h2.write(b"bytes");
     h2.write(b"string");
-    h2.write(&[0xFFu8, 0x01u8]);
-    assert_eq!(h1.finish128(), h2.finish128());
+    h2.write(&test_u8.to_ne_bytes());
+    h2.write(&test_u16.to_ne_bytes());
+    h2.write(&test_u32.to_ne_bytes());
+    h2.write(&test_u64.to_ne_bytes());
+    h2.write(&test_u128.to_ne_bytes());
+    h2.write(&test_usize.to_ne_bytes());
+    h2.write(&test_i8.to_ne_bytes());
+    h2.write(&test_i16.to_ne_bytes());
+    h2.write(&test_i32.to_ne_bytes());
+    h2.write(&test_i64.to_ne_bytes());
+    h2.write(&test_i128.to_ne_bytes());
+    h2.write(&test_isize.to_ne_bytes());
+
+    let h1_hash = h1.finish128();
+    let h2_hash = h2.finish128();
+
+    assert_eq!(h1_hash, h2_hash);
 }

compiler/rustc_data_structures/src/stable_hasher.rs

@@ -5,6 +5,9 @@ use smallvec::SmallVec;
 use std::hash::{BuildHasher, Hash, Hasher};
 use std::mem;
 
+#[cfg(test)]
+mod tests;
+
 /// When hashing something that ends up affecting properties like symbol names,
 /// we want these symbol names to be calculated independently of other factors
 /// like what architecture you're compiling *from*.
@@ -129,7 +132,8 @@ impl Hasher for StableHasher {
     fn write_isize(&mut self, i: isize) {
         // Always treat isize as i64 so we get the same results on 32 and 64 bit
         // platforms. This is important for symbol hashes when cross compiling,
-        // for example.
+        // for example. Sign extending here is preferable as it means that the
+        // same negative number hashes the same on both 32 and 64 bit platforms.
         self.state.write_i64((i as i64).to_le());
     }
 }

compiler/rustc_data_structures/src/stable_hasher/tests.rs

@@ -0,0 +1,73 @@
+use super::*;
+
+// The tests below compare the computed hashes to particular expected values
+// in order to test that we produce the same results on different platforms,
+// regardless of endianness and `usize` and `isize` size differences (this
+// of course assumes we run these tests on platforms that differ in those
+// ways). The expected values depend on the hashing algorithm used, so they
+// need to be updated whenever StableHasher changes its hashing algorithm.
+
+#[test]
+fn test_hash_integers() {
+    // Test that integers are handled consistently across platforms.
+    let test_u8 = 0xAB_u8;
+    let test_u16 = 0xFFEE_u16;
+    let test_u32 = 0x445577AA_u32;
+    let test_u64 = 0x01234567_13243546_u64;
+    let test_u128 = 0x22114433_66557788_99AACCBB_EEDDFF77_u128;
+    let test_usize = 0xD0C0B0A0_usize;
+
+    let test_i8 = -100_i8;
+    let test_i16 = -200_i16;
+    let test_i32 = -300_i32;
+    let test_i64 = -400_i64;
+    let test_i128 = -500_i128;
+    let test_isize = -600_isize;
+
+    let mut h = StableHasher::new();
+    test_u8.hash(&mut h);
+    test_u16.hash(&mut h);
+    test_u32.hash(&mut h);
+    test_u64.hash(&mut h);
+    test_u128.hash(&mut h);
+    test_usize.hash(&mut h);
+    test_i8.hash(&mut h);
+    test_i16.hash(&mut h);
+    test_i32.hash(&mut h);
+    test_i64.hash(&mut h);
+    test_i128.hash(&mut h);
+    test_isize.hash(&mut h);
+
+    // This depends on the hashing algorithm. See note at top of file.
+    let expected = (2736651863462566372, 8121090595289675650);
+
+    assert_eq!(h.finalize(), expected);
+}
+
+#[test]
+fn test_hash_usize() {
+    // Test that usize specifically is handled consistently across platforms.
+    let test_usize = 0xABCDEF01_usize;
+
+    let mut h = StableHasher::new();
+    test_usize.hash(&mut h);
+
+    // This depends on the hashing algorithm. See note at top of file.
+    let expected = (5798740672699530587, 11186240177685111648);
+
+    assert_eq!(h.finalize(), expected);
+}
+
+#[test]
+fn test_hash_isize() {
+    // Test that isize specifically is handled consistently across platforms.
+    let test_isize = -7_isize;
+
+    let mut h = StableHasher::new();
+    test_isize.hash(&mut h);
+
+    // This depends on the hashing algorithm. See note at top of file.
+    let expected = (14721296605626097289, 11385941877786388409);
+
+    assert_eq!(h.finalize(), expected);
+}

compiler/rustc_middle/src/ty/error.rs

@@ -232,7 +232,7 @@ impl<'tcx> ty::TyS<'tcx> {
                 let n = tcx.lift(&n).unwrap();
                 match n.try_eval_usize(tcx, ty::ParamEnv::empty()) {
                     _ if t.is_simple_ty() => format!("array `{}`", self).into(),
-                    Some(n) => format!("array of {} element{} ", n, pluralize!(n)).into(),
+                    Some(n) => format!("array of {} element{}", n, pluralize!(n)).into(),
                     None => "array".into(),
                 }
             }

compiler/rustc_mir/src/const_eval/eval_queries.rs

@@ -14,7 +14,7 @@ use rustc_middle::ty::print::with_no_trimmed_paths;
 use rustc_middle::ty::{self, subst::Subst, TyCtxt};
 use rustc_span::source_map::Span;
 use rustc_target::abi::{Abi, LayoutOf};
-use std::convert::TryInto;
+use std::convert::{TryFrom, TryInto};
 
 pub fn note_on_undefined_behavior_error() -> &'static str {
     "The rules on what exactly is undefined behavior aren't clear, \
@@ -148,10 +148,10 @@ pub(super) fn op_to_const<'tcx>(
         Scalar::Raw { data, .. } => {
             assert!(mplace.layout.is_zst());
             assert_eq!(
-                data,
-                mplace.layout.align.abi.bytes().into(),
-                "this MPlaceTy must come from `try_as_mplace` being used on a zst, so we know what
-                 value this integer address must have",
+                u64::try_from(data).unwrap() % mplace.layout.align.abi.bytes(),
+                0,
+                "this MPlaceTy must come from a validated constant, thus we can assume the \
+                alignment is correct",
             );
             ConstValue::Scalar(Scalar::zst())
         }

compiler/rustc_mir/src/transform/check_const_item_mutation.rs

@@ -60,11 +60,15 @@ impl<'a, 'tcx> Visitor<'tcx> for ConstMutationChecker<'a, 'tcx> {
             // so emitting a lint would be redundant.
             if !lhs.projection.is_empty() {
                 if let Some(def_id) = self.is_const_item(lhs.local) {
-                    self.lint_const_item_usage(def_id, loc, |lint| {
-                        let mut lint = lint.build("attempting to modify a `const` item");
-                        lint.note("each usage of a `const` item creates a new temporary - the original `const` item will not be modified");
-                        lint
-                    })
+                    // Don't lint on writes through a pointer
+                    // (e.g. `unsafe { *FOO = 0; *BAR.field = 1; }`)
+                    if !matches!(lhs.projection.last(), Some(PlaceElem::Deref)) {
+                        self.lint_const_item_usage(def_id, loc, |lint| {
+                            let mut lint = lint.build("attempting to modify a `const` item");
+                            lint.note("each usage of a `const` item creates a new temporary - the original `const` item will not be modified");
+                            lint
+                        })
+                    }
                 }
             }
             // We are looking for MIR of the form:

compiler/rustc_mir/src/transform/promote_consts.rs

@@ -137,7 +137,7 @@ fn args_required_const(tcx: TyCtxt<'_>, def_id: DefId) -> Option<Vec<usize>> {
             LitKind::Int(a, _) => {
                 ret.push(a as usize);
             }
-            _ => return None,
+            _ => bug!("invalid arg index"),
         }
     }
     Some(ret)