bjoernager/rust
bjoernager/rust
1
Fork 0
Code Activity

mv compiler to compiler/

Browse source
This commit is contained in:
mark 2020-08-27 22:58:48 -05:00 • committed by Vadim Petrochenkov
parent db534b3ac2
commit 9e5f7d5631
1686 changed files with 941 additions and 1051 deletions
Download patch file Download diff file Expand all files Collapse all files

95
compiler/rustc_codegen_llvm/src/debuginfo/create_scope_map.rs Normal file
View file

@ -0,0 +1,95 @@
use super::metadata::{file_metadata, UNKNOWN_COLUMN_NUMBER, UNKNOWN_LINE_NUMBER};
use super::utils::DIB;
use rustc_codegen_ssa::mir::debuginfo::{DebugScope, FunctionDebugContext};
use rustc_codegen_ssa::traits::*;
use crate::common::CodegenCx;
use crate::llvm;
use crate::llvm::debuginfo::{DIScope, DISubprogram};
use rustc_middle::mir::{Body, SourceScope};
use rustc_session::config::DebugInfo;
use rustc_index::bit_set::BitSet;
use rustc_index::vec::Idx;
/// Produces DIScope DIEs for each MIR Scope which has variables defined in it.
pub fn compute_mir_scopes(
cx: &CodegenCx<'ll, '_>,
mir: &Body<'_>,
fn_metadata: &'ll DISubprogram,
debug_context: &mut FunctionDebugContext<&'ll DIScope>,
) {
// Find all the scopes with variables defined in them.
let mut has_variables = BitSet::new_empty(mir.source_scopes.len());
// Only consider variables when they're going to be emitted.
// FIXME(eddyb) don't even allocate `has_variables` otherwise.
if cx.sess().opts.debuginfo == DebugInfo::Full {
// FIXME(eddyb) take into account that arguments always have debuginfo,
// irrespective of their name (assuming full debuginfo is enabled).
// NOTE(eddyb) actually, on second thought, those are always in the
// function scope, which always exists.
for var_debug_info in &mir.var_debug_info {
has_variables.insert(var_debug_info.source_info.scope);
}
}
// Instantiate all scopes.
for idx in 0..mir.source_scopes.len() {
let scope = SourceScope::new(idx);
make_mir_scope(cx, &mir, fn_metadata, &has_variables, debug_context, scope);
}
}
fn make_mir_scope(
cx: &CodegenCx<'ll, '_>,
mir: &Body<'_>,
fn_metadata: &'ll DISubprogram,
has_variables: &BitSet<SourceScope>,
debug_context: &mut FunctionDebugContext<&'ll DISubprogram>,
scope: SourceScope,
) {
if debug_context.scopes[scope].is_valid() {
return;
}
let scope_data = &mir.source_scopes[scope];
let parent_scope = if let Some(parent) = scope_data.parent_scope {
make_mir_scope(cx, mir, fn_metadata, has_variables, debug_context, parent);
debug_context.scopes[parent]
} else {
// The root is the function itself.
let loc = cx.lookup_debug_loc(mir.span.lo());
debug_context.scopes[scope] = DebugScope {
scope_metadata: Some(fn_metadata),
file_start_pos: loc.file.start_pos,
file_end_pos: loc.file.end_pos,
};
return;
};
if !has_variables.contains(scope) {
// Do not create a DIScope if there are no variables
// defined in this MIR Scope, to avoid debuginfo bloat.
debug_context.scopes[scope] = parent_scope;
return;
}
let loc = cx.lookup_debug_loc(scope_data.span.lo());
let file_metadata = file_metadata(cx, &loc.file, debug_context.defining_crate);
let scope_metadata = unsafe {
Some(llvm::LLVMRustDIBuilderCreateLexicalBlock(
DIB(cx),
parent_scope.scope_metadata.unwrap(),
file_metadata,
loc.line.unwrap_or(UNKNOWN_LINE_NUMBER),
loc.col.unwrap_or(UNKNOWN_COLUMN_NUMBER),
))
};
debug_context.scopes[scope] = DebugScope {
scope_metadata,
file_start_pos: loc.file.start_pos,
file_end_pos: loc.file.end_pos,
};
}

179
compiler/rustc_codegen_llvm/src/debuginfo/doc.rs Normal file
View file

@ -0,0 +1,179 @@
//! # Debug Info Module
//!
//! This module serves the purpose of generating debug symbols. We use LLVM's
//! [source level debugging](https://llvm.org/docs/SourceLevelDebugging.html)
//! features for generating the debug information. The general principle is
//! this:
//!
//! Given the right metadata in the LLVM IR, the LLVM code generator is able to
//! create DWARF debug symbols for the given code. The
//! [metadata](https://llvm.org/docs/LangRef.html#metadata-type) is structured
//! much like DWARF *debugging information entries* (DIE), representing type
//! information such as datatype layout, function signatures, block layout,
//! variable location and scope information, etc. It is the purpose of this
//! module to generate correct metadata and insert it into the LLVM IR.
//!
//! As the exact format of metadata trees may change between different LLVM
//! versions, we now use LLVM
//! [DIBuilder](https://llvm.org/docs/doxygen/html/classllvm_1_1DIBuilder.html)
//! to create metadata where possible. This will hopefully ease the adaption of
//! this module to future LLVM versions.
//!
//! The public API of the module is a set of functions that will insert the
//! correct metadata into the LLVM IR when called with the right parameters.
//! The module is thus driven from an outside client with functions like
//! `debuginfo::create_local_var_metadata(bx: block, local: &ast::local)`.
//!
//! Internally the module will try to reuse already created metadata by
//! utilizing a cache. The way to get a shared metadata node when needed is
//! thus to just call the corresponding function in this module:
//!
//! let file_metadata = file_metadata(crate_context, path);
//!
//! The function will take care of probing the cache for an existing node for
//! that exact file path.
//!
//! All private state used by the module is stored within either the
//! CrateDebugContext struct (owned by the CodegenCx) or the
//! FunctionDebugContext (owned by the FunctionCx).
//!
//! This file consists of three conceptual sections:
//! 1. The public interface of the module
//! 2. Module-internal metadata creation functions
//! 3. Minor utility functions
//!
//!
//! ## Recursive Types
//!
//! Some kinds of types, such as structs and enums can be recursive. That means
//! that the type definition of some type X refers to some other type which in
//! turn (transitively) refers to X. This introduces cycles into the type
//! referral graph. A naive algorithm doing an on-demand, depth-first traversal
//! of this graph when describing types, can get trapped in an endless loop
//! when it reaches such a cycle.
//!
//! For example, the following simple type for a singly-linked list...
//!
//! ```
//! struct List {
//! value: i32,
//! tail: Option<Box<List>>,
//! }
//! ```
//!
//! will generate the following callstack with a naive DFS algorithm:
//!
//! ```
//! describe(t = List)
//! describe(t = i32)
//! describe(t = Option<Box<List>>)
//! describe(t = Box<List>)
//! describe(t = List) // at the beginning again...
//! ...
//! ```
//!
//! To break cycles like these, we use "forward declarations". That is, when
//! the algorithm encounters a possibly recursive type (any struct or enum), it
//! immediately creates a type description node and inserts it into the cache
//! *before* describing the members of the type. This type description is just
//! a stub (as type members are not described and added to it yet) but it
//! allows the algorithm to already refer to the type. After the stub is
//! inserted into the cache, the algorithm continues as before. If it now
//! encounters a recursive reference, it will hit the cache and does not try to
//! describe the type anew.
//!
//! This behavior is encapsulated in the 'RecursiveTypeDescription' enum,
//! which represents a kind of continuation, storing all state needed to
//! continue traversal at the type members after the type has been registered
//! with the cache. (This implementation approach might be a tad over-
//! engineered and may change in the future)
//!
//!
//! ## Source Locations and Line Information
//!
//! In addition to data type descriptions the debugging information must also
//! allow to map machine code locations back to source code locations in order
//! to be useful. This functionality is also handled in this module. The
//! following functions allow to control source mappings:
//!
//! + set_source_location()
//! + clear_source_location()
//! + start_emitting_source_locations()
//!
//! `set_source_location()` allows to set the current source location. All IR
//! instructions created after a call to this function will be linked to the
//! given source location, until another location is specified with
//! `set_source_location()` or the source location is cleared with
//! `clear_source_location()`. In the later case, subsequent IR instruction
//! will not be linked to any source location. As you can see, this is a
//! stateful API (mimicking the one in LLVM), so be careful with source
//! locations set by previous calls. It's probably best to not rely on any
//! specific state being present at a given point in code.
//!
//! One topic that deserves some extra attention is *function prologues*. At
//! the beginning of a function's machine code there are typically a few
//! instructions for loading argument values into allocas and checking if
//! there's enough stack space for the function to execute. This *prologue* is
//! not visible in the source code and LLVM puts a special PROLOGUE END marker
//! into the line table at the first non-prologue instruction of the function.
//! In order to find out where the prologue ends, LLVM looks for the first
//! instruction in the function body that is linked to a source location. So,
//! when generating prologue instructions we have to make sure that we don't
//! emit source location information until the 'real' function body begins. For
//! this reason, source location emission is disabled by default for any new
//! function being codegened and is only activated after a call to the third
//! function from the list above, `start_emitting_source_locations()`. This
//! function should be called right before regularly starting to codegen the
//! top-level block of the given function.
//!
//! There is one exception to the above rule: `llvm.dbg.declare` instruction
//! must be linked to the source location of the variable being declared. For
//! function parameters these `llvm.dbg.declare` instructions typically occur
//! in the middle of the prologue, however, they are ignored by LLVM's prologue
//! detection. The `create_argument_metadata()` and related functions take care
//! of linking the `llvm.dbg.declare` instructions to the correct source
//! locations even while source location emission is still disabled, so there
//! is no need to do anything special with source location handling here.
//!
//! ## Unique Type Identification
//!
//! In order for link-time optimization to work properly, LLVM needs a unique
//! type identifier that tells it across compilation units which types are the
//! same as others. This type identifier is created by
//! `TypeMap::get_unique_type_id_of_type()` using the following algorithm:
//!
//! (1) Primitive types have their name as ID
//! (2) Structs, enums and traits have a multipart identifier
//!
//! (1) The first part is the SVH (strict version hash) of the crate they
//! were originally defined in
//!
//! (2) The second part is the ast::NodeId of the definition in their
//! original crate
//!
//! (3) The final part is a concatenation of the type IDs of their concrete
//! type arguments if they are generic types.
//!
//! (3) Tuple-, pointer and function types are structurally identified, which
//! means that they are equivalent if their component types are equivalent
//! (i.e., (i32, i32) is the same regardless in which crate it is used).
//!
//! This algorithm also provides a stable ID for types that are defined in one
//! crate but instantiated from metadata within another crate. We just have to
//! take care to always map crate and `NodeId`s back to the original crate
//! context.
//!
//! As a side-effect these unique type IDs also help to solve a problem arising
//! from lifetime parameters. Since lifetime parameters are completely omitted
//! in debuginfo, more than one `Ty` instance may map to the same debuginfo
//! type metadata, that is, some struct `Struct<'a>` may have N instantiations
//! with different concrete substitutions for `'a`, and thus there will be N
//! `Ty` instances for the type `Struct<'a>` even though it is not generic
//! otherwise. Unfortunately this means that we cannot use `ty::type_id()` as
//! cheap identifier for type metadata -- we have done this in the past, but it
//! led to unnecessary metadata duplication in the best case and LLVM
//! assertions in the worst. However, the unique type ID as described above
//! *can* be used as identifier. Since it is comparatively expensive to
//! construct, though, `ty::type_id()` is still used additionally as an
//! optimization for cases where the exact same type has been seen before
//! (which is most of the time).

71
compiler/rustc_codegen_llvm/src/debuginfo/gdb.rs Normal file
View file

@ -0,0 +1,71 @@
// .debug_gdb_scripts binary section.
use crate::llvm;
use crate::builder::Builder;
use crate::common::CodegenCx;
use crate::value::Value;
use rustc_codegen_ssa::traits::*;
use rustc_middle::bug;
use rustc_session::config::DebugInfo;
use rustc_span::symbol::sym;
/// Inserts a side-effect free instruction sequence that makes sure that the
/// .debug_gdb_scripts global is referenced, so it isn't removed by the linker.
pub fn insert_reference_to_gdb_debug_scripts_section_global(bx: &mut Builder<'_, '_, '_>) {
if needs_gdb_debug_scripts_section(bx) {
let gdb_debug_scripts_section = get_or_insert_gdb_debug_scripts_section_global(bx);
// Load just the first byte as that's all that's necessary to force
// LLVM to keep around the reference to the global.
let indices = [bx.const_i32(0), bx.const_i32(0)];
let element = bx.inbounds_gep(gdb_debug_scripts_section, &indices);
let volative_load_instruction = bx.volatile_load(element);
unsafe {
llvm::LLVMSetAlignment(volative_load_instruction, 1);
}
}
}
/// Allocates the global variable responsible for the .debug_gdb_scripts binary
/// section.
pub fn get_or_insert_gdb_debug_scripts_section_global(cx: &CodegenCx<'ll, '_>) -> &'ll Value {
let c_section_var_name = "__rustc_debug_gdb_scripts_section__\0";
let section_var_name = &c_section_var_name[..c_section_var_name.len() - 1];
let section_var =
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 section_contents = b"\x01gdb_load_rust_pretty_printers.py\0";
unsafe {
let llvm_type = cx.type_array(cx.type_i8(), section_contents.len() as u64);
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, 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);
llvm::LLVMRustSetLinkage(section_var, llvm::Linkage::LinkOnceODRLinkage);
// This should make sure that the whole section is not larger than
// the string it contains. Otherwise we get a warning from GDB.
llvm::LLVMSetAlignment(section_var, 1);
section_var
}
})
}
pub fn needs_gdb_debug_scripts_section(cx: &CodegenCx<'_, '_>) -> bool {
let omit_gdb_pretty_printer_section = cx
.tcx
.sess
.contains_name(&cx.tcx.hir().krate_attrs(), sym::omit_gdb_pretty_printer_section);
!omit_gdb_pretty_printer_section
&& cx.sess().opts.debuginfo != DebugInfo::None
&& cx.sess().target.target.options.emit_debug_gdb_scripts
}

2585
compiler/rustc_codegen_llvm/src/debuginfo/metadata.rs Normal file
View file

File diff suppressed because it is too large Load diff

563
compiler/rustc_codegen_llvm/src/debuginfo/mod.rs Normal file
View file

@ -0,0 +1,563 @@
// See doc.rs for documentation.
mod doc;
use rustc_codegen_ssa::mir::debuginfo::VariableKind::*;
use self::metadata::{file_metadata, type_metadata, TypeMap, UNKNOWN_LINE_NUMBER};
use self::namespace::mangled_name_of_instance;
use self::type_names::compute_debuginfo_type_name;
use self::utils::{create_DIArray, is_node_local_to_unit, DIB};
use crate::abi::FnAbi;
use crate::builder::Builder;
use crate::common::CodegenCx;
use crate::llvm;
use crate::llvm::debuginfo::{
DIArray, DIBuilder, DIFile, DIFlags, DILexicalBlock, DISPFlags, DIScope, DIType, DIVariable,
};
use crate::value::Value;
use rustc_codegen_ssa::debuginfo::type_names;
use rustc_codegen_ssa::mir::debuginfo::{DebugScope, FunctionDebugContext, VariableKind};
use rustc_codegen_ssa::traits::*;
use rustc_data_structures::fx::{FxHashMap, FxHashSet};
use rustc_hir::def_id::{CrateNum, DefId, DefIdMap, LOCAL_CRATE};
use rustc_index::vec::IndexVec;
use rustc_middle::mir;
use rustc_middle::ty::layout::HasTyCtxt;
use rustc_middle::ty::subst::{GenericArgKind, SubstsRef};
use rustc_middle::ty::{self, Instance, ParamEnv, Ty, TypeFoldable};
use rustc_session::config::{self, DebugInfo};
use rustc_span::symbol::Symbol;
use rustc_span::{self, BytePos, Span};
use rustc_target::abi::{LayoutOf, Primitive, Size};
use libc::c_uint;
use smallvec::SmallVec;
use std::cell::RefCell;
use tracing::debug;
mod create_scope_map;
pub mod gdb;
pub mod metadata;
mod namespace;
mod source_loc;
mod utils;
pub use self::create_scope_map::compute_mir_scopes;
pub use self::metadata::create_global_var_metadata;
pub use self::metadata::extend_scope_to_file;
#[allow(non_upper_case_globals)]
const DW_TAG_auto_variable: c_uint = 0x100;
#[allow(non_upper_case_globals)]
const DW_TAG_arg_variable: c_uint = 0x101;
/// A context object for maintaining all state needed by the debuginfo module.
pub struct CrateDebugContext<'a, 'tcx> {
llcontext: &'a llvm::Context,
llmod: &'a llvm::Module,
builder: &'a mut DIBuilder<'a>,
created_files: RefCell<FxHashMap<(Option<String>, Option<String>), &'a DIFile>>,
created_enum_disr_types: RefCell<FxHashMap<(DefId, Primitive), &'a DIType>>,
type_map: RefCell<TypeMap<'a, 'tcx>>,
namespace_map: RefCell<DefIdMap<&'a DIScope>>,
// This collection is used to assert that composite types (structs, enums,
// ...) have their members only set once:
composite_types_completed: RefCell<FxHashSet<&'a DIType>>,
}
impl Drop for CrateDebugContext<'a, 'tcx> {
fn drop(&mut self) {
unsafe {
llvm::LLVMRustDIBuilderDispose(&mut *(self.builder as *mut _));
}
}
}
impl<'a, 'tcx> CrateDebugContext<'a, 'tcx> {
pub fn new(llmod: &'a llvm::Module) -> Self {
debug!("CrateDebugContext::new");
let builder = unsafe { llvm::LLVMRustDIBuilderCreate(llmod) };
// DIBuilder inherits context from the module, so we'd better use the same one
let llcontext = unsafe { llvm::LLVMGetModuleContext(llmod) };
CrateDebugContext {
llcontext,
llmod,
builder,
created_files: Default::default(),
created_enum_disr_types: Default::default(),
type_map: Default::default(),
namespace_map: RefCell::new(Default::default()),
composite_types_completed: Default::default(),
}
}
}
/// Creates any deferred debug metadata nodes
pub fn finalize(cx: &CodegenCx<'_, '_>) {
if cx.dbg_cx.is_none() {
return;
}
debug!("finalize");
if gdb::needs_gdb_debug_scripts_section(cx) {
// Add a .debug_gdb_scripts section to this compile-unit. This will
// cause GDB to try and load the gdb_load_rust_pretty_printers.py file,
// which activates the Rust pretty printers for binary this section is
// contained in.
gdb::get_or_insert_gdb_debug_scripts_section_global(cx);
}
unsafe {
llvm::LLVMRustDIBuilderFinalize(DIB(cx));
// Debuginfo generation in LLVM by default uses a higher
// version of dwarf than macOS currently understands. We can
// instruct LLVM to emit an older version of dwarf, however,
// for macOS to understand. For more info see #11352
// This can be overridden using --llvm-opts -dwarf-version,N.
// Android has the same issue (#22398)
if cx.sess().target.target.options.is_like_osx
|| cx.sess().target.target.options.is_like_android
{
llvm::LLVMRustAddModuleFlag(cx.llmod, "Dwarf Version\0".as_ptr().cast(), 2)
}
// Indicate that we want CodeView debug information on MSVC
if cx.sess().target.target.options.is_like_msvc {
llvm::LLVMRustAddModuleFlag(cx.llmod, "CodeView\0".as_ptr().cast(), 1)
}
// Prevent bitcode readers from deleting the debug info.
let ptr = "Debug Info Version\0".as_ptr();
llvm::LLVMRustAddModuleFlag(cx.llmod, ptr.cast(), llvm::LLVMRustDebugMetadataVersion());
};
}
impl DebugInfoBuilderMethods for Builder<'a, 'll, 'tcx> {
// FIXME(eddyb) find a common convention for all of the debuginfo-related
// names (choose between `dbg`, `debug`, `debuginfo`, `debug_info` etc.).
fn dbg_var_addr(
&mut self,
dbg_var: &'ll DIVariable,
scope_metadata: &'ll DIScope,
variable_alloca: Self::Value,
direct_offset: Size,
indirect_offsets: &[Size],
span: Span,
) {
let cx = self.cx();
// Convert the direct and indirect offsets to address ops.
// FIXME(eddyb) use `const`s instead of getting the values via FFI,
// the values should match the ones in the DWARF standard anyway.
let op_deref = || unsafe { llvm::LLVMRustDIBuilderCreateOpDeref() };
let op_plus_uconst = || unsafe { llvm::LLVMRustDIBuilderCreateOpPlusUconst() };
let mut addr_ops = SmallVec::<[_; 8]>::new();
if direct_offset.bytes() > 0 {
addr_ops.push(op_plus_uconst());
addr_ops.push(direct_offset.bytes() as i64);
}
for &offset in indirect_offsets {
addr_ops.push(op_deref());
if offset.bytes() > 0 {
addr_ops.push(op_plus_uconst());
addr_ops.push(offset.bytes() as i64);
}
}
// FIXME(eddyb) maybe this information could be extracted from `dbg_var`,
// to avoid having to pass it down in both places?
// NB: `var` doesn't seem to know about the column, so that's a limitation.
let dbg_loc = cx.create_debug_loc(scope_metadata, span);
unsafe {
// FIXME(eddyb) replace `llvm.dbg.declare` with `llvm.dbg.addr`.
llvm::LLVMRustDIBuilderInsertDeclareAtEnd(
DIB(cx),
variable_alloca,
dbg_var,
addr_ops.as_ptr(),
addr_ops.len() as c_uint,
dbg_loc,
self.llbb(),
);
}
}
fn set_source_location(&mut self, scope: &'ll DIScope, span: Span) {
debug!("set_source_location: {}", self.sess().source_map().span_to_string(span));
let dbg_loc = self.cx().create_debug_loc(scope, span);
unsafe {
llvm::LLVMSetCurrentDebugLocation(self.llbuilder, dbg_loc);
}
}
fn insert_reference_to_gdb_debug_scripts_section_global(&mut self) {
gdb::insert_reference_to_gdb_debug_scripts_section_global(self)
}
fn set_var_name(&mut self, value: &'ll Value, name: &str) {
// Avoid wasting time if LLVM value names aren't even enabled.
if self.sess().fewer_names() {
return;
}
// Only function parameters and instructions are local to a function,
// don't change the name of anything else (e.g. globals).
let param_or_inst = unsafe {
llvm::LLVMIsAArgument(value).is_some() || llvm::LLVMIsAInstruction(value).is_some()
};
if !param_or_inst {
return;
}
// Avoid replacing the name if it already exists.
// While we could combine the names somehow, it'd
// get noisy quick, and the usefulness is dubious.
if llvm::get_value_name(value).is_empty() {
llvm::set_value_name(value, name.as_bytes());
}
}
}
impl DebugInfoMethods<'tcx> for CodegenCx<'ll, 'tcx> {
fn create_function_debug_context(
&self,
instance: Instance<'tcx>,
fn_abi: &FnAbi<'tcx, Ty<'tcx>>,
llfn: &'ll Value,
mir: &mir::Body<'_>,
) -> Option<FunctionDebugContext<&'ll DIScope>> {
if self.sess().opts.debuginfo == DebugInfo::None {
return None;
}
let span = mir.span;
// This can be the case for functions inlined from another crate
if span.is_dummy() {
// FIXME(simulacrum): Probably can't happen; remove.
return None;
}
let def_id = instance.def_id();
let containing_scope = get_containing_scope(self, instance);
let loc = self.lookup_debug_loc(span.lo());
let file_metadata = file_metadata(self, &loc.file, def_id.krate);
let function_type_metadata = unsafe {
let fn_signature = get_function_signature(self, fn_abi);
llvm::LLVMRustDIBuilderCreateSubroutineType(DIB(self), fn_signature)
};
// Find the enclosing function, in case this is a closure.
let def_key = self.tcx().def_key(def_id);
let mut name = def_key.disambiguated_data.data.to_string();
let enclosing_fn_def_id = self.tcx().closure_base_def_id(def_id);
// Get_template_parameters() will append a `<...>` clause to the function
// name if necessary.
let generics = self.tcx().generics_of(enclosing_fn_def_id);
let substs = instance.substs.truncate_to(self.tcx(), generics);
let template_parameters = get_template_parameters(self, &generics, substs, &mut name);
let linkage_name = &mangled_name_of_instance(self, instance).name;
// Omit the linkage_name if it is the same as subprogram name.
let linkage_name = if &name == linkage_name { "" } else { linkage_name };
// FIXME(eddyb) does this need to be separate from `loc.line` for some reason?
let scope_line = loc.line;
let mut flags = DIFlags::FlagPrototyped;
if fn_abi.ret.layout.abi.is_uninhabited() {
flags |= DIFlags::FlagNoReturn;
}
let mut spflags = DISPFlags::SPFlagDefinition;
if is_node_local_to_unit(self, def_id) {
spflags |= DISPFlags::SPFlagLocalToUnit;
}
if self.sess().opts.optimize != config::OptLevel::No {
spflags |= DISPFlags::SPFlagOptimized;
}
if let Some((id, _)) = self.tcx.entry_fn(LOCAL_CRATE) {
if id.to_def_id() == def_id {
spflags |= DISPFlags::SPFlagMainSubprogram;
}
}
let fn_metadata = unsafe {
llvm::LLVMRustDIBuilderCreateFunction(
DIB(self),
containing_scope,
name.as_ptr().cast(),
name.len(),
linkage_name.as_ptr().cast(),
linkage_name.len(),
file_metadata,
loc.line.unwrap_or(UNKNOWN_LINE_NUMBER),
function_type_metadata,
scope_line.unwrap_or(UNKNOWN_LINE_NUMBER),
flags,
spflags,
llfn,
template_parameters,
None,
)
};
// Initialize fn debug context (including scopes).
// FIXME(eddyb) figure out a way to not need `Option` for `scope_metadata`.
let null_scope = DebugScope {
scope_metadata: None,
file_start_pos: BytePos(0),
file_end_pos: BytePos(0),
};
let mut fn_debug_context = FunctionDebugContext {
scopes: IndexVec::from_elem(null_scope, &mir.source_scopes),
defining_crate: def_id.krate,
};
// Fill in all the scopes, with the information from the MIR body.
compute_mir_scopes(self, mir, fn_metadata, &mut fn_debug_context);
return Some(fn_debug_context);
fn get_function_signature<'ll, 'tcx>(
cx: &CodegenCx<'ll, 'tcx>,
fn_abi: &FnAbi<'tcx, Ty<'tcx>>,
) -> &'ll DIArray {
if cx.sess().opts.debuginfo == DebugInfo::Limited {
return create_DIArray(DIB(cx), &[]);
}
let mut signature = Vec::with_capacity(fn_abi.args.len() + 1);
// Return type -- llvm::DIBuilder wants this at index 0
signature.push(if fn_abi.ret.is_ignore() {
None
} else {
Some(type_metadata(cx, fn_abi.ret.layout.ty, rustc_span::DUMMY_SP))
});
// Arguments types
if cx.sess().target.target.options.is_like_msvc {
// FIXME(#42800):
// There is a bug in MSDIA that leads to a crash when it encounters
// a fixed-size array of `u8` or something zero-sized in a
// function-type (see #40477).
// As a workaround, we replace those fixed-size arrays with a
// pointer-type. So a function `fn foo(a: u8, b: [u8; 4])` would
// appear as `fn foo(a: u8, b: *const u8)` in debuginfo,
// and a function `fn bar(x: [(); 7])` as `fn bar(x: *const ())`.
// This transformed type is wrong, but these function types are
// already inaccurate due to ABI adjustments (see #42800).
signature.extend(fn_abi.args.iter().map(|arg| {
let t = arg.layout.ty;
let t = match t.kind {
ty::Array(ct, _)
if (ct == cx.tcx.types.u8) || cx.layout_of(ct).is_zst() =>
{
cx.tcx.mk_imm_ptr(ct)
}
_ => t,
};
Some(type_metadata(cx, t, rustc_span::DUMMY_SP))
}));
} else {
signature.extend(
fn_abi
.args
.iter()
.map(|arg| Some(type_metadata(cx, arg.layout.ty, rustc_span::DUMMY_SP))),
);
}
create_DIArray(DIB(cx), &signature[..])
}
fn get_template_parameters<'ll, 'tcx>(
cx: &CodegenCx<'ll, 'tcx>,
generics: &ty::Generics,
substs: SubstsRef<'tcx>,
name_to_append_suffix_to: &mut String,
) -> &'ll DIArray {
if substs.types().next().is_none() {
return create_DIArray(DIB(cx), &[]);
}
name_to_append_suffix_to.push('<');
for (i, actual_type) in substs.types().enumerate() {
if i != 0 {
name_to_append_suffix_to.push_str(",");
}
let actual_type =
cx.tcx.normalize_erasing_regions(ParamEnv::reveal_all(), actual_type);
// Add actual type name to <...> clause of function name
let actual_type_name = compute_debuginfo_type_name(cx.tcx(), actual_type, true);
name_to_append_suffix_to.push_str(&actual_type_name[..]);
}
name_to_append_suffix_to.push('>');
// Again, only create type information if full debuginfo is enabled
let template_params: Vec<_> = if cx.sess().opts.debuginfo == DebugInfo::Full {
let names = get_parameter_names(cx, generics);
substs
.iter()
.zip(names)
.filter_map(|(kind, name)| {
if let GenericArgKind::Type(ty) = kind.unpack() {
let actual_type =
cx.tcx.normalize_erasing_regions(ParamEnv::reveal_all(), ty);
let actual_type_metadata =
type_metadata(cx, actual_type, rustc_span::DUMMY_SP);
let name = name.as_str();
Some(unsafe {
Some(llvm::LLVMRustDIBuilderCreateTemplateTypeParameter(
DIB(cx),
None,
name.as_ptr().cast(),
name.len(),
actual_type_metadata,
))
})
} else {
None
}
})
.collect()
} else {
vec![]
};
create_DIArray(DIB(cx), &template_params[..])
}
fn get_parameter_names(cx: &CodegenCx<'_, '_>, generics: &ty::Generics) -> Vec<Symbol> {
let mut names = generics
.parent
.map_or(vec![], |def_id| get_parameter_names(cx, cx.tcx.generics_of(def_id)));
names.extend(generics.params.iter().map(|param| param.name));
names
}
fn get_containing_scope<'ll, 'tcx>(
cx: &CodegenCx<'ll, 'tcx>,
instance: Instance<'tcx>,
) -> &'ll DIScope {
// First, let's see if this is a method within an inherent impl. Because
// if yes, we want to make the result subroutine DIE a child of the
// subroutine's self-type.
let self_type = cx.tcx.impl_of_method(instance.def_id()).and_then(|impl_def_id| {
// If the method does *not* belong to a trait, proceed
if cx.tcx.trait_id_of_impl(impl_def_id).is_none() {
let impl_self_ty = cx.tcx.subst_and_normalize_erasing_regions(
instance.substs,
ty::ParamEnv::reveal_all(),
&cx.tcx.type_of(impl_def_id),
);
// Only "class" methods are generally understood by LLVM,
// so avoid methods on other types (e.g., `<*mut T>::null`).
match impl_self_ty.kind {
ty::Adt(def, ..) if !def.is_box() => {
// Again, only create type information if full debuginfo is enabled
if cx.sess().opts.debuginfo == DebugInfo::Full
&& !impl_self_ty.needs_subst()
{
Some(type_metadata(cx, impl_self_ty, rustc_span::DUMMY_SP))
} else {
Some(namespace::item_namespace(cx, def.did))
}
}
_ => None,
}
} else {
// For trait method impls we still use the "parallel namespace"
// strategy
None
}
});
self_type.unwrap_or_else(|| {
namespace::item_namespace(
cx,
DefId {
krate: instance.def_id().krate,
index: cx
.tcx
.def_key(instance.def_id())
.parent
.expect("get_containing_scope: missing parent?"),
},
)
})
}
}
fn create_vtable_metadata(&self, ty: Ty<'tcx>, vtable: Self::Value) {
metadata::create_vtable_metadata(self, ty, vtable)
}
fn extend_scope_to_file(
&self,
scope_metadata: &'ll DIScope,
file: &rustc_span::SourceFile,
defining_crate: CrateNum,
) -> &'ll DILexicalBlock {
metadata::extend_scope_to_file(&self, scope_metadata, file, defining_crate)
}
fn debuginfo_finalize(&self) {
finalize(self)
}
// FIXME(eddyb) find a common convention for all of the debuginfo-related
// names (choose between `dbg`, `debug`, `debuginfo`, `debug_info` etc.).
fn create_dbg_var(
&self,
dbg_context: &FunctionDebugContext<&'ll DIScope>,
variable_name: Symbol,
variable_type: Ty<'tcx>,
scope_metadata: &'ll DIScope,
variable_kind: VariableKind,
span: Span,
) -> &'ll DIVariable {
let loc = self.lookup_debug_loc(span.lo());
let file_metadata = file_metadata(self, &loc.file, dbg_context.defining_crate);
let type_metadata = type_metadata(self, variable_type, span);
let (argument_index, dwarf_tag) = match variable_kind {
ArgumentVariable(index) => (index as c_uint, DW_TAG_arg_variable),
LocalVariable => (0, DW_TAG_auto_variable),
};
let align = self.align_of(variable_type);
let name = variable_name.as_str();
unsafe {
llvm::LLVMRustDIBuilderCreateVariable(
DIB(self),
dwarf_tag,
scope_metadata,
name.as_ptr().cast(),
name.len(),
file_metadata,
loc.line.unwrap_or(UNKNOWN_LINE_NUMBER),
type_metadata,
true,
DIFlags::FlagZero,
argument_index,
align.bytes() as u32,
)
}
}
}

48
compiler/rustc_codegen_llvm/src/debuginfo/namespace.rs Normal file
View file

@ -0,0 +1,48 @@
// Namespace Handling.
use super::utils::{debug_context, DIB};
use rustc_middle::ty::{self, Instance};
use crate::common::CodegenCx;
use crate::llvm;
use crate::llvm::debuginfo::DIScope;
use rustc_hir::def_id::DefId;
use rustc_hir::definitions::DefPathData;
pub fn mangled_name_of_instance<'a, 'tcx>(
cx: &CodegenCx<'a, 'tcx>,
instance: Instance<'tcx>,
) -> ty::SymbolName<'tcx> {
let tcx = cx.tcx;
tcx.symbol_name(instance)
}
pub fn item_namespace(cx: &CodegenCx<'ll, '_>, def_id: DefId) -> &'ll DIScope {
if let Some(&scope) = debug_context(cx).namespace_map.borrow().get(&def_id) {
return scope;
}
let def_key = cx.tcx.def_key(def_id);
let parent_scope = def_key
.parent
.map(|parent| item_namespace(cx, DefId { krate: def_id.krate, index: parent }));
let namespace_name = match def_key.disambiguated_data.data {
DefPathData::CrateRoot => cx.tcx.crate_name(def_id.krate),
data => data.as_symbol(),
};
let namespace_name = namespace_name.as_str();
let scope = unsafe {
llvm::LLVMRustDIBuilderCreateNameSpace(
DIB(cx),
parent_scope,
namespace_name.as_ptr().cast(),
namespace_name.len(),
false, // ExportSymbols (only relevant for C++ anonymous namespaces)
)
};
debug_context(cx).namespace_map.borrow_mut().insert(def_id, scope);
scope
}

61
compiler/rustc_codegen_llvm/src/debuginfo/source_loc.rs Normal file
View file

@ -0,0 +1,61 @@
use super::metadata::{UNKNOWN_COLUMN_NUMBER, UNKNOWN_LINE_NUMBER};
use super::utils::debug_context;
use crate::common::CodegenCx;
use crate::llvm::debuginfo::DIScope;
use crate::llvm::{self, Value};
use rustc_codegen_ssa::traits::*;
use rustc_data_structures::sync::Lrc;
use rustc_span::{BytePos, Pos, SourceFile, SourceFileAndLine, Span};
/// A source code location used to generate debug information.
pub struct DebugLoc {
/// Information about the original source file.
pub file: Lrc<SourceFile>,
/// The (1-based) line number.
pub line: Option<u32>,
/// The (1-based) column number.
pub col: Option<u32>,
}
impl CodegenCx<'ll, '_> {
/// Looks up debug source information about a `BytePos`.
pub fn lookup_debug_loc(&self, pos: BytePos) -> DebugLoc {
let (file, line, col) = match self.sess().source_map().lookup_line(pos) {
Ok(SourceFileAndLine { sf: file, line }) => {
let line_pos = file.line_begin_pos(pos);
// Use 1-based indexing.
let line = (line + 1) as u32;
let col = (pos - line_pos).to_u32() + 1;
(file, Some(line), Some(col))
}
Err(file) => (file, None, None),
};
// For MSVC, omit the column number.
// Otherwise, emit it. This mimics clang behaviour.
// See discussion in https://github.com/rust-lang/rust/issues/42921
if self.sess().target.target.options.is_like_msvc {
DebugLoc { file, line, col: None }
} else {
DebugLoc { file, line, col }
}
}
pub fn create_debug_loc(&self, scope: &'ll DIScope, span: Span) -> &'ll Value {
let DebugLoc { line, col, .. } = self.lookup_debug_loc(span.lo());
unsafe {
llvm::LLVMRustDIBuilderCreateDebugLocation(
debug_context(self).llcontext,
line.unwrap_or(UNKNOWN_LINE_NUMBER),
col.unwrap_or(UNKNOWN_COLUMN_NUMBER),
scope,
None,
)
}
}
}

43
compiler/rustc_codegen_llvm/src/debuginfo/utils.rs Normal file
View file

@ -0,0 +1,43 @@
// Utility Functions.
use super::namespace::item_namespace;
use super::CrateDebugContext;
use rustc_hir::def_id::DefId;
use rustc_middle::ty::DefIdTree;
use crate::common::CodegenCx;
use crate::llvm;
use crate::llvm::debuginfo::{DIArray, DIBuilder, DIDescriptor, DIScope};
pub fn is_node_local_to_unit(cx: &CodegenCx<'_, '_>, def_id: DefId) -> bool {
// The is_local_to_unit flag indicates whether a function is local to the
// current compilation unit (i.e., if it is *static* in the C-sense). The
// *reachable* set should provide a good approximation of this, as it
// contains everything that might leak out of the current crate (by being
// externally visible or by being inlined into something externally
// visible). It might better to use the `exported_items` set from
// `driver::CrateAnalysis` in the future, but (atm) this set is not
// available in the codegen pass.
!cx.tcx.is_reachable_non_generic(def_id)
}
#[allow(non_snake_case)]
pub fn create_DIArray(builder: &DIBuilder<'ll>, arr: &[Option<&'ll DIDescriptor>]) -> &'ll DIArray {
unsafe { llvm::LLVMRustDIBuilderGetOrCreateArray(builder, arr.as_ptr(), arr.len() as u32) }
}
#[inline]
pub fn debug_context(cx: &'a CodegenCx<'ll, 'tcx>) -> &'a CrateDebugContext<'ll, 'tcx> {
cx.dbg_cx.as_ref().unwrap()
}
#[inline]
#[allow(non_snake_case)]
pub fn DIB(cx: &'a CodegenCx<'ll, '_>) -> &'a DIBuilder<'ll> {
cx.dbg_cx.as_ref().unwrap().builder
}
pub fn get_namespace_for_item(cx: &CodegenCx<'ll, '_>, def_id: DefId) -> &'ll DIScope {
item_namespace(cx, cx.tcx.parent(def_id).expect("get_namespace_for_item: missing parent?"))
}