Rollup merge of #113334 - fmease:revert-lexing-c-str-lits, r=compiler…

…-errors

Revert the lexing of `c"…"` string literals

Fixes \[after beta-backport\] #113235.
Further progress is tracked in #113333.

This PR *manually* reverts parts of #108801 (since a git-revert would've been too coarse-grained & messy)
and git-reverts #111647.

CC `@fee1-dead` (#108801) `@klensy` (#111647)
r? `@compiler-errors`

`@rustbot` label F-c_str_literals beta-nominated

Cargo.lock

@@ -766,6 +766,16 @@ dependencies = [
  "typenum",
 ]
 
+[[package]]
+name = "cstr"
+version = "0.2.8"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c11a39d776a3b35896711da8a04dc1835169dcd36f710878187637314e47941b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+]
+
 [[package]]
 name = "ctrlc"
 version = "3.4.0"
@@ -3016,6 +3026,7 @@ name = "rustc_codegen_llvm"
 version = "0.0.0"
 dependencies = [
  "bitflags 1.3.2",
+ "cstr",
  "libc",
  "measureme",
  "object",

compiler/rustc_codegen_llvm/Cargo.toml

@@ -8,6 +8,7 @@ test = false
 
 [dependencies]
 bitflags = "1.0"
+cstr = "0.2"
 libc = "0.2"
 measureme = "10.0.0"
 object = { version = "0.31.1", default-features = false, features = [

compiler/rustc_codegen_llvm/src/allocator.rs

@@ -77,7 +77,7 @@ pub(crate) unsafe fn codegen(
                 llvm::LLVMRustGetOrInsertFunction(llmod, callee.as_ptr().cast(), callee.len(), ty);
             llvm::LLVMRustSetVisibility(callee, llvm::Visibility::Hidden);
 
-            let llbb = llvm::LLVMAppendBasicBlockInContext(llcx, llfn, c"entry".as_ptr().cast());
+            let llbb = llvm::LLVMAppendBasicBlockInContext(llcx, llfn, "entry\0".as_ptr().cast());
 
             let llbuilder = llvm::LLVMCreateBuilderInContext(llcx);
             llvm::LLVMPositionBuilderAtEnd(llbuilder, llbb);
@@ -129,7 +129,7 @@ pub(crate) unsafe fn codegen(
     attributes::apply_to_llfn(callee, llvm::AttributePlace::Function, &[no_return]);
     llvm::LLVMRustSetVisibility(callee, llvm::Visibility::Hidden);
 
-    let llbb = llvm::LLVMAppendBasicBlockInContext(llcx, llfn, c"entry".as_ptr().cast());
+    let llbb = llvm::LLVMAppendBasicBlockInContext(llcx, llfn, "entry\0".as_ptr().cast());
 
     let llbuilder = llvm::LLVMCreateBuilderInContext(llcx);
     llvm::LLVMPositionBuilderAtEnd(llbuilder, llbb);

compiler/rustc_codegen_llvm/src/back/lto.rs

@@ -601,7 +601,7 @@ pub(crate) fn run_pass_manager(
             llvm::LLVMRustAddModuleFlag(
                 module.module_llvm.llmod(),
                 llvm::LLVMModFlagBehavior::Error,
-                c"LTOPostLink".as_ptr().cast(),
+                "LTOPostLink\0".as_ptr().cast(),
                 1,
             );
         }

compiler/rustc_codegen_llvm/src/back/write.rs

@@ -931,16 +931,16 @@ unsafe fn embed_bitcode(
         let llglobal = llvm::LLVMAddGlobal(
             llmod,
             common::val_ty(llconst),
-            c"rustc.embedded.module".as_ptr().cast(),
+            "rustc.embedded.module\0".as_ptr().cast(),
         );
         llvm::LLVMSetInitializer(llglobal, llconst);
 
         let section = if is_apple {
-            c"__LLVM,__bitcode"
+            "__LLVM,__bitcode\0"
         } else if is_aix {
-            c".ipa"
+            ".ipa\0"
         } else {
-            c".llvmbc"
+            ".llvmbc\0"
         };
         llvm::LLVMSetSection(llglobal, section.as_ptr().cast());
         llvm::LLVMRustSetLinkage(llglobal, llvm::Linkage::PrivateLinkage);
@@ -950,15 +950,15 @@ unsafe fn embed_bitcode(
         let llglobal = llvm::LLVMAddGlobal(
             llmod,
             common::val_ty(llconst),
-            c"rustc.embedded.cmdline".as_ptr().cast(),
+            "rustc.embedded.cmdline\0".as_ptr().cast(),
         );
         llvm::LLVMSetInitializer(llglobal, llconst);
         let section = if is_apple {
-            c"__LLVM,__cmdline"
+            "__LLVM,__cmdline\0"
         } else if is_aix {
-            c".info"
+            ".info\0"
         } else {
-            c".llvmcmd"
+            ".llvmcmd\0"
         };
         llvm::LLVMSetSection(llglobal, section.as_ptr().cast());
         llvm::LLVMRustSetLinkage(llglobal, llvm::Linkage::PrivateLinkage);

compiler/rustc_codegen_llvm/src/base.rs

@@ -19,6 +19,8 @@ use crate::context::CodegenCx;
 use crate::llvm;
 use crate::value::Value;
 
+use cstr::cstr;
+
 use rustc_codegen_ssa::base::maybe_create_entry_wrapper;
 use rustc_codegen_ssa::mono_item::MonoItemExt;
 use rustc_codegen_ssa::traits::*;
@@ -108,11 +110,11 @@ pub fn compile_codegen_unit(tcx: TyCtxt<'_>, cgu_name: Symbol) -> (ModuleCodegen
 
             // Create the llvm.used and llvm.compiler.used variables.
             if !cx.used_statics.borrow().is_empty() {
-                cx.create_used_variable_impl(c"llvm.used", &*cx.used_statics.borrow());
+                cx.create_used_variable_impl(cstr!("llvm.used"), &*cx.used_statics.borrow());
             }
             if !cx.compiler_used_statics.borrow().is_empty() {
                 cx.create_used_variable_impl(
-                    c"llvm.compiler.used",
+                    cstr!("llvm.compiler.used"),
                     &*cx.compiler_used_statics.borrow(),
                 );
             }

compiler/rustc_codegen_llvm/src/builder.rs

@@ -6,6 +6,7 @@ use crate::llvm::{self, AtomicOrdering, AtomicRmwBinOp, BasicBlock, False, True}
 use crate::type_::Type;
 use crate::type_of::LayoutLlvmExt;
 use crate::value::Value;
+use cstr::cstr;
 use libc::{c_char, c_uint};
 use rustc_codegen_ssa::common::{IntPredicate, RealPredicate, SynchronizationScope, TypeKind};
 use rustc_codegen_ssa::mir::operand::{OperandRef, OperandValue};
@@ -25,6 +26,7 @@ use rustc_target::abi::{self, call::FnAbi, Align, Size, WrappingRange};
 use rustc_target::spec::{HasTargetSpec, SanitizerSet, Target};
 use smallvec::SmallVec;
 use std::borrow::Cow;
+use std::ffi::CStr;
 use std::iter;
 use std::ops::Deref;
 use std::ptr;
@@ -44,10 +46,13 @@ impl Drop for Builder<'_, '_, '_> {
     }
 }
 
+// FIXME(eddyb) use a checked constructor when they become `const fn`.
+const EMPTY_C_STR: &CStr = unsafe { CStr::from_bytes_with_nul_unchecked(b"\0") };
+
 /// Empty string, to be used where LLVM expects an instruction name, indicating
 /// that the instruction is to be left unnamed (i.e. numbered, in textual IR).
 // FIXME(eddyb) pass `&CStr` directly to FFI once it's a thin pointer.
-const UNNAMED: *const c_char = c"".as_ptr();
+const UNNAMED: *const c_char = EMPTY_C_STR.as_ptr();
 
 impl<'ll, 'tcx> BackendTypes for Builder<'_, 'll, 'tcx> {
     type Value = <CodegenCx<'ll, 'tcx> as BackendTypes>::Value;
@@ -1002,13 +1007,14 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
     }
 
     fn cleanup_pad(&mut self, parent: Option<&'ll Value>, args: &[&'ll Value]) -> Funclet<'ll> {
+        let name = cstr!("cleanuppad");
         let ret = unsafe {
             llvm::LLVMBuildCleanupPad(
                 self.llbuilder,
                 parent,
                 args.as_ptr(),
                 args.len() as c_uint,
-                c"cleanuppad".as_ptr(),
+                name.as_ptr(),
             )
         };
         Funclet::new(ret.expect("LLVM does not have support for cleanuppad"))
@@ -1022,13 +1028,14 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
     }
 
     fn catch_pad(&mut self, parent: &'ll Value, args: &[&'ll Value]) -> Funclet<'ll> {
+        let name = cstr!("catchpad");
         let ret = unsafe {
             llvm::LLVMBuildCatchPad(
                 self.llbuilder,
                 parent,
                 args.as_ptr(),
                 args.len() as c_uint,
-                c"catchpad".as_ptr(),
+                name.as_ptr(),
             )
         };
         Funclet::new(ret.expect("LLVM does not have support for catchpad"))
@@ -1040,13 +1047,14 @@ impl<'a, 'll, 'tcx> BuilderMethods<'a, 'tcx> for Builder<'a, 'll, 'tcx> {
         unwind: Option<&'ll BasicBlock>,
         handlers: &[&'ll BasicBlock],
     ) -> &'ll Value {
+        let name = cstr!("catchswitch");
         let ret = unsafe {
             llvm::LLVMBuildCatchSwitch(
                 self.llbuilder,
                 parent,
                 unwind,
                 handlers.len() as c_uint,
-                c"catchswitch".as_ptr(),
+                name.as_ptr(),
             )
         };
         let ret = ret.expect("LLVM does not have support for catchswitch");

compiler/rustc_codegen_llvm/src/consts.rs

@@ -8,6 +8,7 @@ use crate::llvm::{self, True};
 use crate::type_::Type;
 use crate::type_of::LayoutLlvmExt;
 use crate::value::Value;
+use cstr::cstr;
 use rustc_codegen_ssa::traits::*;
 use rustc_hir::def_id::DefId;
 use rustc_middle::middle::codegen_fn_attrs::{CodegenFnAttrFlags, CodegenFnAttrs};
@@ -481,9 +482,9 @@ impl<'ll> StaticMethods for CodegenCx<'ll, '_> {
                             .all(|&byte| byte == 0);
 
                     let sect_name = if all_bytes_are_zero {
-                        c"__DATA,__thread_bss"
+                        cstr!("__DATA,__thread_bss")
                     } else {
-                        c"__DATA,__thread_data"
+                        cstr!("__DATA,__thread_data")
                     };
                     llvm::LLVMSetSection(g, sect_name.as_ptr());
                 }
@@ -512,7 +513,7 @@ impl<'ll> StaticMethods for CodegenCx<'ll, '_> {
                     let val = llvm::LLVMMetadataAsValue(self.llcx, meta);
                     llvm::LLVMAddNamedMetadataOperand(
                         self.llmod,
-                        c"wasm.custom_sections".as_ptr().cast(),
+                        "wasm.custom_sections\0".as_ptr().cast(),
                         val,
                     );
                 }

compiler/rustc_codegen_llvm/src/context.rs

@@ -8,6 +8,7 @@ use crate::llvm_util;
 use crate::type_::Type;
 use crate::value::Value;
 
+use cstr::cstr;
 use rustc_codegen_ssa::base::{wants_msvc_seh, wants_wasm_eh};
 use rustc_codegen_ssa::traits::*;
 use rustc_data_structures::base_n;
@@ -223,42 +224,36 @@ pub unsafe fn create_module<'ll>(
     // If skipping the PLT is enabled, we need to add some module metadata
     // to ensure intrinsic calls don't use it.
     if !sess.needs_plt() {
-        llvm::LLVMRustAddModuleFlag(
-            llmod,
-            llvm::LLVMModFlagBehavior::Warning,
-            c"RtLibUseGOT".as_ptr().cast(),
-            1,
-        );
+        let avoid_plt = "RtLibUseGOT\0".as_ptr().cast();
+        llvm::LLVMRustAddModuleFlag(llmod, llvm::LLVMModFlagBehavior::Warning, avoid_plt, 1);
     }
 
     // Enable canonical jump tables if CFI is enabled. (See https://reviews.llvm.org/D65629.)
     if sess.is_sanitizer_cfi_canonical_jump_tables_enabled() && sess.is_sanitizer_cfi_enabled() {
+        let canonical_jump_tables = "CFI Canonical Jump Tables\0".as_ptr().cast();
         llvm::LLVMRustAddModuleFlag(
             llmod,
             llvm::LLVMModFlagBehavior::Override,
-            c"CFI Canonical Jump Tables".as_ptr().cast(),
+            canonical_jump_tables,
             1,
         );
     }
 
     // Enable LTO unit splitting if specified or if CFI is enabled. (See https://reviews.llvm.org/D53891.)
     if sess.is_split_lto_unit_enabled() || sess.is_sanitizer_cfi_enabled() {
+        let enable_split_lto_unit = "EnableSplitLTOUnit\0".as_ptr().cast();
         llvm::LLVMRustAddModuleFlag(
             llmod,
             llvm::LLVMModFlagBehavior::Override,
-            c"EnableSplitLTOUnit".as_ptr().cast(),
+            enable_split_lto_unit,
             1,
         );
     }
 
     // Add "kcfi" module flag if KCFI is enabled. (See https://reviews.llvm.org/D119296.)
     if sess.is_sanitizer_kcfi_enabled() {
-        llvm::LLVMRustAddModuleFlag(
-            llmod,
-            llvm::LLVMModFlagBehavior::Override,
-            c"kcfi".as_ptr().cast(),
-            1,
-        );
+        let kcfi = "kcfi\0".as_ptr().cast();
+        llvm::LLVMRustAddModuleFlag(llmod, llvm::LLVMModFlagBehavior::Override, kcfi, 1);
     }
 
     // Control Flow Guard is currently only supported by the MSVC linker on Windows.
@@ -270,7 +265,7 @@ pub unsafe fn create_module<'ll>(
                 llvm::LLVMRustAddModuleFlag(
                     llmod,
                     llvm::LLVMModFlagBehavior::Warning,
-                    c"cfguard".as_ptr() as *const _,
+                    "cfguard\0".as_ptr() as *const _,
                     1,
                 )
             }
@@ -279,7 +274,7 @@ pub unsafe fn create_module<'ll>(
                 llvm::LLVMRustAddModuleFlag(
                     llmod,
                     llvm::LLVMModFlagBehavior::Warning,
-                    c"cfguard".as_ptr() as *const _,
+                    "cfguard\0".as_ptr() as *const _,
                     2,
                 )
             }
@@ -297,26 +292,26 @@ pub unsafe fn create_module<'ll>(
             llvm::LLVMRustAddModuleFlag(
                 llmod,
                 behavior,
-                c"branch-target-enforcement".as_ptr().cast(),
+                "branch-target-enforcement\0".as_ptr().cast(),
                 bti.into(),
             );
             llvm::LLVMRustAddModuleFlag(
                 llmod,
                 behavior,
-                c"sign-return-address".as_ptr().cast(),
+                "sign-return-address\0".as_ptr().cast(),
                 pac_ret.is_some().into(),
             );
             let pac_opts = pac_ret.unwrap_or(PacRet { leaf: false, key: PAuthKey::A });
             llvm::LLVMRustAddModuleFlag(
                 llmod,
                 behavior,
-                c"sign-return-address-all".as_ptr().cast(),
+                "sign-return-address-all\0".as_ptr().cast(),
                 pac_opts.leaf.into(),
             );
             llvm::LLVMRustAddModuleFlag(
                 llmod,
                 behavior,
-                c"sign-return-address-with-bkey".as_ptr().cast(),
+                "sign-return-address-with-bkey\0".as_ptr().cast(),
                 u32::from(pac_opts.key == PAuthKey::B),
             );
         } else {
@@ -332,15 +327,15 @@ pub unsafe fn create_module<'ll>(
         llvm::LLVMRustAddModuleFlag(
             llmod,
             llvm::LLVMModFlagBehavior::Override,
-            c"cf-protection-branch".as_ptr().cast(),
+            "cf-protection-branch\0".as_ptr().cast(),
             1,
         )
     }
     if let CFProtection::Return | CFProtection::Full = sess.opts.unstable_opts.cf_protection {
         llvm::LLVMRustAddModuleFlag(
             llmod,
             llvm::LLVMModFlagBehavior::Override,
-            c"cf-protection-return".as_ptr().cast(),
+            "cf-protection-return\0".as_ptr().cast(),
             1,
         )
     }
@@ -349,7 +344,7 @@ pub unsafe fn create_module<'ll>(
         llvm::LLVMRustAddModuleFlag(
             llmod,
             llvm::LLVMModFlagBehavior::Error,
-            c"Virtual Function Elim".as_ptr().cast(),
+            "Virtual Function Elim\0".as_ptr().cast(),
             1,
         );
     }
@@ -481,13 +476,14 @@ impl<'ll, 'tcx> CodegenCx<'ll, 'tcx> {
     }
 
     pub(crate) fn create_used_variable_impl(&self, name: &'static CStr, values: &[&'ll Value]) {
+        let section = cstr!("llvm.metadata");
         let array = self.const_array(self.type_ptr_to(self.type_i8()), values);
 
         unsafe {
             let g = llvm::LLVMAddGlobal(self.llmod, self.val_ty(array), name.as_ptr());
             llvm::LLVMSetInitializer(g, array);
             llvm::LLVMRustSetLinkage(g, llvm::Linkage::AppendingLinkage);
-            llvm::LLVMSetSection(g, c"llvm.metadata".as_ptr());
+            llvm::LLVMSetSection(g, section.as_ptr());
         }
     }
 }

compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs

@@ -38,6 +38,7 @@ pub fn get_or_insert_gdb_debug_scripts_section_global<'ll>(cx: &CodegenCx<'ll, '
         unsafe { llvm::LLVMGetNamedGlobal(cx.llmod, c_section_var_name.as_ptr().cast()) };
 
     section_var.unwrap_or_else(|| {
+        let section_name = b".debug_gdb_scripts\0";
         let mut section_contents = Vec::new();
 
         // Add the pretty printers for the standard library first.
@@ -70,7 +71,7 @@ pub fn get_or_insert_gdb_debug_scripts_section_global<'ll>(cx: &CodegenCx<'ll, '
             let section_var = cx
                 .define_global(section_var_name, llvm_type)
                 .unwrap_or_else(|| bug!("symbol `{}` is already defined", section_var_name));
-            llvm::LLVMSetSection(section_var, c".debug_gdb_scripts".as_ptr().cast());
+            llvm::LLVMSetSection(section_var, section_name.as_ptr().cast());
             llvm::LLVMSetInitializer(section_var, cx.const_bytes(section_contents));
             llvm::LLVMSetGlobalConstant(section_var, llvm::True);
             llvm::LLVMSetUnnamedAddress(section_var, llvm::UnnamedAddr::Global);