Auto merge of rust-lang#86225 - Mark-Simulacrum:beta-next, r=Mark-Sim…

…ulacrum

[beta] backports

* Disable the machine outliner by default rust-lang#86020
* Fix incorrect gating of nonterminals in key-value attributes rust-lang#85445
* Build crtbegin.o/crtend.o from source code rust-lang#85395
* Bring back x86_64-sun-solaris target to rustup rust-lang#85252
* Preserve SyntaxContext for invalid/dummy spans in crate metadata rust-lang#85211
* [beta] backport: Remove unsound TrustedRandomAccess implementations rust-lang#86222

r? `@Mark-Simulacrum`

compiler/rustc_codegen_llvm/src/llvm_util.rs

@@ -1,5 +1,5 @@
 use crate::back::write::create_informational_target_machine;
-use crate::llvm;
+use crate::{llvm, llvm_util};
 use libc::c_int;
 use rustc_codegen_ssa::target_features::supported_target_features;
 use rustc_data_structures::fx::FxHashSet;
@@ -84,6 +84,17 @@ unsafe fn configure_llvm(sess: &Session) {
         if !sess.opts.debugging_opts.no_generate_arange_section {
             add("-generate-arange-section", false);
         }
+
+        // FIXME(nagisa): disable the machine outliner by default in LLVM versions 11, where it was
+        // introduced and up.
+        //
+        // This should remain in place until https://reviews.llvm.org/D103167 is fixed. If LLVM
+        // has been upgraded since, consider adjusting the version check below to contain an upper
+        // bound.
+        if llvm_util::get_version() >= (11, 0, 0) {
+            add("-enable-machine-outliner=never", false);
+        }
+
         match sess.opts.debugging_opts.merge_functions.unwrap_or(sess.target.merge_functions) {
             MergeFunctions::Disabled | MergeFunctions::Trampolines => {}
             MergeFunctions::Aliases => {

compiler/rustc_metadata/src/rmeta/decoder.rs

@@ -406,17 +406,17 @@ impl<'a, 'tcx> Decodable<DecodeContext<'a, 'tcx>> for ExpnId {
 
 impl<'a, 'tcx> Decodable<DecodeContext<'a, 'tcx>> for Span {
     fn decode(decoder: &mut DecodeContext<'a, 'tcx>) -> Result<Span, String> {
+        let ctxt = SyntaxContext::decode(decoder)?;
         let tag = u8::decode(decoder)?;
 
-        if tag == TAG_INVALID_SPAN {
-            return Ok(DUMMY_SP);
+        if tag == TAG_PARTIAL_SPAN {
+            return Ok(DUMMY_SP.with_ctxt(ctxt));
         }
 
         debug_assert!(tag == TAG_VALID_SPAN_LOCAL || tag == TAG_VALID_SPAN_FOREIGN);
 
         let lo = BytePos::decode(decoder)?;
         let len = BytePos::decode(decoder)?;
-        let ctxt = SyntaxContext::decode(decoder)?;
         let hi = lo + len;
 
         let sess = if let Some(sess) = decoder.sess {

compiler/rustc_metadata/src/rmeta/encoder.rs

@@ -184,11 +184,48 @@ impl<'a, 'tcx> Encodable<EncodeContext<'a, 'tcx>> for ExpnId {
 
 impl<'a, 'tcx> Encodable<EncodeContext<'a, 'tcx>> for Span {
     fn encode(&self, s: &mut EncodeContext<'a, 'tcx>) -> opaque::EncodeResult {
-        if *self == rustc_span::DUMMY_SP {
-            return TAG_INVALID_SPAN.encode(s);
+        let span = self.data();
+
+        // Don't serialize any `SyntaxContext`s from a proc-macro crate,
+        // since we don't load proc-macro dependencies during serialization.
+        // This means that any hygiene information from macros used *within*
+        // a proc-macro crate (e.g. invoking a macro that expands to a proc-macro
+        // definition) will be lost.
+        //
+        // This can show up in two ways:
+        //
+        // 1. Any hygiene information associated with identifier of
+        // a proc macro (e.g. `#[proc_macro] pub fn $name`) will be lost.
+        // Since proc-macros can only be invoked from a different crate,
+        // real code should never need to care about this.
+        //
+        // 2. Using `Span::def_site` or `Span::mixed_site` will not
+        // include any hygiene information associated with the definition
+        // site. This means that a proc-macro cannot emit a `$crate`
+        // identifier which resolves to one of its dependencies,
+        // which also should never come up in practice.
+        //
+        // Additionally, this affects `Span::parent`, and any other
+        // span inspection APIs that would otherwise allow traversing
+        // the `SyntaxContexts` associated with a span.
+        //
+        // None of these user-visible effects should result in any
+        // cross-crate inconsistencies (getting one behavior in the same
+        // crate, and a different behavior in another crate) due to the
+        // limited surface that proc-macros can expose.
+        //
+        // IMPORTANT: If this is ever changed, be sure to update
+        // `rustc_span::hygiene::raw_encode_expn_id` to handle
+        // encoding `ExpnData` for proc-macro crates.
+        if s.is_proc_macro {
+            SyntaxContext::root().encode(s)?;
+        } else {
+            span.ctxt.encode(s)?;
         }
 
-        let span = self.data();
+        if self.is_dummy() {
+            return TAG_PARTIAL_SPAN.encode(s);
+        }
 
         // The Span infrastructure should make sure that this invariant holds:
         debug_assert!(span.lo <= span.hi);
@@ -203,7 +240,7 @@ impl<'a, 'tcx> Encodable<EncodeContext<'a, 'tcx>> for Span {
         if !s.source_file_cache.0.contains(span.hi) {
             // Unfortunately, macro expansion still sometimes generates Spans
             // that malformed in this way.
-            return TAG_INVALID_SPAN.encode(s);
+            return TAG_PARTIAL_SPAN.encode(s);
         }
 
         let source_files = s.required_source_files.as_mut().expect("Already encoded SourceMap!");
@@ -259,43 +296,6 @@ impl<'a, 'tcx> Encodable<EncodeContext<'a, 'tcx>> for Span {
         let len = hi - lo;
         len.encode(s)?;
 
-        // Don't serialize any `SyntaxContext`s from a proc-macro crate,
-        // since we don't load proc-macro dependencies during serialization.
-        // This means that any hygiene information from macros used *within*
-        // a proc-macro crate (e.g. invoking a macro that expands to a proc-macro
-        // definition) will be lost.
-        //
-        // This can show up in two ways:
-        //
-        // 1. Any hygiene information associated with identifier of
-        // a proc macro (e.g. `#[proc_macro] pub fn $name`) will be lost.
-        // Since proc-macros can only be invoked from a different crate,
-        // real code should never need to care about this.
-        //
-        // 2. Using `Span::def_site` or `Span::mixed_site` will not
-        // include any hygiene information associated with the definition
-        // site. This means that a proc-macro cannot emit a `$crate`
-        // identifier which resolves to one of its dependencies,
-        // which also should never come up in practice.
-        //
-        // Additionally, this affects `Span::parent`, and any other
-        // span inspection APIs that would otherwise allow traversing
-        // the `SyntaxContexts` associated with a span.
-        //
-        // None of these user-visible effects should result in any
-        // cross-crate inconsistencies (getting one behavior in the same
-        // crate, and a different behavior in another crate) due to the
-        // limited surface that proc-macros can expose.
-        //
-        // IMPORTANT: If this is ever changed, be sure to update
-        // `rustc_span::hygiene::raw_encode_expn_id` to handle
-        // encoding `ExpnData` for proc-macro crates.
-        if s.is_proc_macro {
-            SyntaxContext::root().encode(s)?;
-        } else {
-            span.ctxt.encode(s)?;
-        }
-
         if tag == TAG_VALID_SPAN_FOREIGN {
             // This needs to be two lines to avoid holding the `s.source_file_cache`
             // while calling `cnum.encode(s)`

compiler/rustc_metadata/src/rmeta/mod.rs

@@ -450,4 +450,4 @@ struct GeneratorData<'tcx> {
 // Tags used for encoding Spans:
 const TAG_VALID_SPAN_LOCAL: u8 = 0;
 const TAG_VALID_SPAN_FOREIGN: u8 = 1;
-const TAG_INVALID_SPAN: u8 = 2;
+const TAG_PARTIAL_SPAN: u8 = 2;

compiler/rustc_parse/src/parser/mod.rs

@@ -1077,8 +1077,11 @@ impl<'a> Parser<'a> {
                     let span = expr.span;
 
                     match &expr.kind {
-                        // Not gated to supporte things like `doc = $expr` that work on stable.
-                        _ if is_interpolated_expr => {}
+                        // Not gated to support things like `doc = $expr` that work on stable.
+                        // Do not gate in `capture_cfg` mode, since we flatten all nontemrinals
+                        // before parsing. `capture_cfg` mode is only used to reparse existing
+                        // tokens, so the gating will be performed by the initial parse
+                        _ if is_interpolated_expr || self.capture_cfg => {}
                         ExprKind::Lit(lit) if lit.kind.is_unsuffixed() => {}
                         _ => self.sess.gated_spans.gate(sym::extended_key_value_attributes, span),
                     }

library/alloc/src/collections/vec_deque/into_iter.rs

@@ -1,5 +1,5 @@
 use core::fmt;
-use core::iter::{FusedIterator, TrustedLen, TrustedRandomAccess};
+use core::iter::{FusedIterator, TrustedLen};
 
 use super::VecDeque;
 
@@ -36,22 +36,6 @@ impl<T> Iterator for IntoIter<T> {
         let len = self.inner.len();
         (len, Some(len))
     }
-
-    #[inline]
-    unsafe fn __iterator_get_unchecked(&mut self, idx: usize) -> Self::Item
-    where
-        Self: TrustedRandomAccess,
-    {
-        // Safety: The TrustedRandomAccess contract requires that callers only pass an index
-        // that is in bounds.
-        // Additionally Self: TrustedRandomAccess is only implemented for T: Copy which means even
-        // multiple repeated reads of the same index would be safe and the
-        // values are !Drop, thus won't suffer from double drops.
-        unsafe {
-            let idx = self.inner.wrap_add(self.inner.tail, idx);
-            self.inner.buffer_read(idx)
-        }
-    }
 }
 
 #[stable(feature = "rust1", since = "1.0.0")]
@@ -74,14 +58,3 @@ impl<T> FusedIterator for IntoIter<T> {}
 
 #[unstable(feature = "trusted_len", issue = "37572")]
 unsafe impl<T> TrustedLen for IntoIter<T> {}
-
-#[doc(hidden)]
-#[unstable(feature = "trusted_random_access", issue = "none")]
-// T: Copy as approximation for !Drop since get_unchecked does not update the pointers
-// and thus we can't implement drop-handling
-unsafe impl<T> TrustedRandomAccess for IntoIter<T>
-where
-    T: Copy,
-{
-    const MAY_HAVE_SIDE_EFFECT: bool = false;
-}

library/core/src/array/iter.rs

@@ -2,7 +2,7 @@
 
 use crate::{
     fmt,
-    iter::{self, ExactSizeIterator, FusedIterator, TrustedLen, TrustedRandomAccess},
+    iter::{self, ExactSizeIterator, FusedIterator, TrustedLen},
     mem::{self, MaybeUninit},
     ops::Range,
     ptr,
@@ -130,18 +130,6 @@ impl<T, const N: usize> Iterator for IntoIter<T, N> {
     fn last(mut self) -> Option<Self::Item> {
         self.next_back()
     }
-
-    #[inline]
-    unsafe fn __iterator_get_unchecked(&mut self, idx: usize) -> Self::Item
-    where
-        Self: TrustedRandomAccess,
-    {
-        // SAFETY: Callers are only allowed to pass an index that is in bounds
-        // Additionally Self: TrustedRandomAccess is only implemented for T: Copy which means even
-        // multiple repeated reads of the same index would be safe and the
-        // values aree !Drop, thus won't suffer from double drops.
-        unsafe { self.data.get_unchecked(self.alive.start + idx).assume_init_read() }
-    }
 }
 
 #[stable(feature = "array_value_iter_impls", since = "1.40.0")]
@@ -196,17 +184,6 @@ impl<T, const N: usize> FusedIterator for IntoIter<T, N> {}
 #[stable(feature = "array_value_iter_impls", since = "1.40.0")]
 unsafe impl<T, const N: usize> TrustedLen for IntoIter<T, N> {}
 
-#[doc(hidden)]
-#[unstable(feature = "trusted_random_access", issue = "none")]
-// T: Copy as approximation for !Drop since get_unchecked does not update the pointers
-// and thus we can't implement drop-handling
-unsafe impl<T, const N: usize> TrustedRandomAccess for IntoIter<T, N>
-where
-    T: Copy,
-{
-    const MAY_HAVE_SIDE_EFFECT: bool = false;
-}
-
 #[stable(feature = "array_value_iter_impls", since = "1.40.0")]
 impl<T: Clone, const N: usize> Clone for IntoIter<T, N> {
     fn clone(&self) -> Self {

src/bootstrap/builder.rs

@@ -369,7 +369,8 @@ impl<'a> Builder<'a> {
                 tool::Rustfmt,
                 tool::Miri,
                 tool::CargoMiri,
-                native::Lld
+                native::Lld,
+                native::CrtBeginEnd
             ),
             Kind::Check | Kind::Clippy { .. } | Kind::Fix | Kind::Format => describe!(
                 check::Std,

src/bootstrap/compile.rs

@@ -199,8 +199,9 @@ fn copy_self_contained_objects(
                 DependencyType::TargetSelfContained,
             );
         }
+        let crt_path = builder.ensure(native::CrtBeginEnd { target });
         for &obj in &["crtbegin.o", "crtbeginS.o", "crtend.o", "crtendS.o"] {
-            let src = compiler_file(builder, builder.cc(target), target, obj);
+            let src = crt_path.join(obj);
             let target = libdir_self_contained.join(obj);
             builder.copy(&src, &target);
             target_deps.push((target, DependencyType::TargetSelfContained));

src/bootstrap/native.rs

@@ -858,3 +858,69 @@ impl HashStamp {
         fs::write(&self.path, self.hash.as_deref().unwrap_or(b""))
     }
 }
+
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
+pub struct CrtBeginEnd {
+    pub target: TargetSelection,
+}
+
+impl Step for CrtBeginEnd {
+    type Output = PathBuf;
+
+    fn should_run(run: ShouldRun<'_>) -> ShouldRun<'_> {
+        run.path("src/llvm-project/compiler-rt/lib/crt")
+    }
+
+    fn make_run(run: RunConfig<'_>) {
+        run.builder.ensure(CrtBeginEnd { target: run.target });
+    }
+
+    /// Build crtbegin.o/crtend.o for musl target.
+    fn run(self, builder: &Builder<'_>) -> Self::Output {
+        let out_dir = builder.native_dir(self.target).join("crt");
+
+        if builder.config.dry_run {
+            return out_dir;
+        }
+
+        let crtbegin_src = builder.src.join("src/llvm-project/compiler-rt/lib/crt/crtbegin.c");
+        let crtend_src = builder.src.join("src/llvm-project/compiler-rt/lib/crt/crtend.c");
+        if up_to_date(&crtbegin_src, &out_dir.join("crtbegin.o"))
+            && up_to_date(&crtend_src, &out_dir.join("crtendS.o"))
+        {
+            return out_dir;
+        }
+
+        builder.info("Building crtbegin.o and crtend.o");
+        t!(fs::create_dir_all(&out_dir));
+
+        let mut cfg = cc::Build::new();
+
+        if let Some(ar) = builder.ar(self.target) {
+            cfg.archiver(ar);
+        }
+        cfg.compiler(builder.cc(self.target));
+        cfg.cargo_metadata(false)
+            .out_dir(&out_dir)
+            .target(&self.target.triple)
+            .host(&builder.config.build.triple)
+            .warnings(false)
+            .debug(false)
+            .opt_level(3)
+            .file(crtbegin_src)
+            .file(crtend_src);
+
+        // Those flags are defined in src/llvm-project/compiler-rt/lib/crt/CMakeLists.txt
+        // Currently only consumer of those objects is musl, which use .init_array/.fini_array
+        // instead of .ctors/.dtors
+        cfg.flag("-std=c11")
+            .define("CRT_HAS_INITFINI_ARRAY", None)
+            .define("EH_USE_FRAME_REGISTRY", None);
+
+        cfg.compile("crt");
+
+        t!(fs::copy(out_dir.join("crtbegin.o"), out_dir.join("crtbeginS.o")));
+        t!(fs::copy(out_dir.join("crtend.o"), out_dir.join("crtendS.o")));
+        out_dir
+    }
+}

src/ci/docker/host-x86_64/dist-various-1/Dockerfile

@@ -144,7 +144,11 @@ ENV TARGETS=$TARGETS,armv7a-none-eabi
 # riscv targets currently do not need a C compiler, as compiler_builtins
 # doesn't currently have it enabled, and the riscv gcc compiler is not
 # installed.
-ENV CC_mipsel_unknown_linux_musl=mipsel-openwrt-linux-gcc \
+ENV CFLAGS_armv5te_unknown_linux_musleabi="-march=armv5te -marm -mfloat-abi=soft" \
+    CFLAGS_arm_unknown_linux_musleabi="-march=armv6 -marm" \
+    CFLAGS_arm_unknown_linux_musleabihf="-march=armv6 -marm -mfpu=vfp" \
+    CFLAGS_armv7_unknown_linux_musleabihf="-march=armv7-a" \
+    CC_mipsel_unknown_linux_musl=mipsel-openwrt-linux-gcc \
     CC_mips_unknown_linux_musl=mips-openwrt-linux-gcc \
     CC_mips64el_unknown_linux_muslabi64=mips64el-linux-gnuabi64-gcc \
     CC_mips64_unknown_linux_muslabi64=mips64-linux-gnuabi64-gcc \

src/ci/docker/host-x86_64/dist-various-2/Dockerfile

@@ -42,6 +42,9 @@ ENV \
     AR_x86_64_pc_solaris=x86_64-pc-solaris2.10-ar \
     CC_x86_64_pc_solaris=x86_64-pc-solaris2.10-gcc \
     CXX_x86_64_pc_solaris=x86_64-pc-solaris2.10-g++ \
+    AR_x86_64_sun_solaris=x86_64-sun-solaris2.10-ar \
+    CC_x86_64_sun_solaris=x86_64-sun-solaris2.10-gcc \
+    CXX_x86_64_sun_solaris=x86_64-sun-solaris2.10-g++ \
     CC_armv7_unknown_linux_gnueabi=arm-linux-gnueabi-gcc-8 \
     CXX_armv7_unknown_linux_gnueabi=arm-linux-gnueabi-g++-8 \
     AR_x86_64_fortanix_unknown_sgx=ar \
@@ -68,8 +71,10 @@ COPY host-x86_64/dist-various-2/shared.sh /tmp/
 COPY host-x86_64/dist-various-2/build-fuchsia-toolchain.sh /tmp/
 RUN /tmp/build-fuchsia-toolchain.sh
 COPY host-x86_64/dist-various-2/build-solaris-toolchain.sh /tmp/
-RUN /tmp/build-solaris-toolchain.sh x86_64  amd64   solaris-i386
-RUN /tmp/build-solaris-toolchain.sh sparcv9 sparcv9 solaris-sparc
+RUN /tmp/build-solaris-toolchain.sh x86_64  amd64   solaris-i386  pc
+# Build deprecated target 'x86_64-sun-solaris2.10' until removed
+RUN /tmp/build-solaris-toolchain.sh x86_64  amd64   solaris-i386  sun
+RUN /tmp/build-solaris-toolchain.sh sparcv9 sparcv9 solaris-sparc sun
 COPY host-x86_64/dist-various-2/build-x86_64-fortanix-unknown-sgx-toolchain.sh /tmp/
 RUN /tmp/build-x86_64-fortanix-unknown-sgx-toolchain.sh
 
@@ -99,6 +104,7 @@ ENV TARGETS=$TARGETS,wasm32-unknown-unknown
 ENV TARGETS=$TARGETS,wasm32-wasi
 ENV TARGETS=$TARGETS,sparcv9-sun-solaris
 ENV TARGETS=$TARGETS,x86_64-pc-solaris
+ENV TARGETS=$TARGETS,x86_64-sun-solaris
 ENV TARGETS=$TARGETS,x86_64-unknown-linux-gnux32
 ENV TARGETS=$TARGETS,x86_64-fortanix-unknown-sgx
 ENV TARGETS=$TARGETS,nvptx64-nvidia-cuda