DWARF parsing and writing support using LLVM (#2520)

This imports LLVM code for DWARF handling. That code has the
Apache 2 license like us. It's also the same code used to
emit DWARF in the common toolchain, so it seems like a safe choice.

This adds two passes: --dwarfdump which runs the same code LLVM
runs for llvm-dwarfdump. This shows we can parse it ok, and will
be useful for debugging. And --dwarfupdate writes out the DWARF
sections (unchanged from what we read, so it just roundtrips - for
updating we need #2515).

This puts LLVM in thirdparty which is added here.

All the LLVM code is behind USE_LLVM_DWARF, which is on
by default, but off in JS for now, as it increases code size by 20%.

This current approach imports the LLVM files directly. This is not
how they are intended to be used, so it required a bunch of
local changes - more than I expected actually, for the platform-specific
stuff. For now this seems to work, so it may be good enough, but
in the long term we may want to switch to linking against libllvm.
A downside to doing that is that binaryen users would need to
have an LLVM build, and even in the waterfall builds we'd have a
problem - while we ship LLVM there anyhow, we constantly update
it, which means that binaryen would need to be on latest llvm all
the time too (which otherwise, given DWARF is quite stable, we
might not need to constantly update).

An even larger issue is that as I did this work I learned about how
DWARF works in LLVM, and while the reading code is easy to
reuse, the writing code is trickier. The main code path is heavily
integrated with the MC layer, which we don't have - we might want
to create a "fake MC layer" for that, but it sounds hard. Instead,
there is the YAML path which is used mostly for testing, and which
can convert DWARF to and from YAML and from binary. Using
the non-YAML parts there, we can convert binary DWARF to
the YAML layer's nice Info data, then convert that to binary. This
works, however, this is not the path LLVM uses normally, and it
supports only some basic DWARF sections - I had to add ranges
support, in fact. So if we need more complex things, we may end
up needing to use the MC layer approach, or consider some other
DWARF library. However, hopefully that should not affect the core
binaryen code which just calls a library for DWARF stuff.

Helps #2400

5 files changed, 547 insertions, 0 deletions

diff --git a/third_party/llvm-project/include/llvm-c/DataTypes.h b/third_party/llvm-project/include/llvm-c/DataTypes.h
new file mode 100644
index 000000000..893b22b49
--- /dev/null
+++ b/third_party/llvm-project/include/llvm-c/DataTypes.h
@@ -0,0 +1,90 @@
+/*===-- include/llvm-c/DataTypes.h - Define fixed size types ------*- C -*-===*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*|
+|*                                                                            *|
+|* This file contains definitions to figure out the size of _HOST_ data types.*|
+|* This file is important because different host OS's define different macros,*|
+|* which makes portability tough.  This file exports the following            *|
+|* definitions:                                                               *|
+|*                                                                            *|
+|*   [u]int(32|64)_t : typedefs for signed and unsigned 32/64 bit system types*|
+|*   [U]INT(8|16|32|64)_(MIN|MAX) : Constants for the min and max values.     *|
+|*                                                                            *|
+|* No library is required when using these functions.                         *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*/
+
+/* Please leave this file C-compatible. */
+
+#ifndef LLVM_C_DATATYPES_H
+#define LLVM_C_DATATYPES_H
+
+#ifdef __cplusplus
+#include <cmath>
+#else
+#include <math.h>
+#endif
+
+#include <inttypes.h>
+#include <stdint.h>
+
+#ifndef _MSC_VER
+
+#if !defined(UINT32_MAX)
+# error "The standard header <cstdint> is not C++11 compliant. Must #define "\
+        "__STDC_LIMIT_MACROS before #including llvm-c/DataTypes.h"
+#endif
+
+#if !defined(UINT32_C)
+# error "The standard header <cstdint> is not C++11 compliant. Must #define "\
+        "__STDC_CONSTANT_MACROS before #including llvm-c/DataTypes.h"
+#endif
+
+/* Note that <inttypes.h> includes <stdint.h>, if this is a C99 system. */
+#include <sys/types.h>
+
+#ifdef _AIX
+// GCC is strict about defining large constants: they must have LL modifier.
+#undef INT64_MAX
+#undef INT64_MIN
+#endif
+
+#else /* _MSC_VER */
+#ifdef __cplusplus
+#include <cstddef>
+#include <cstdlib>
+#else
+#include <stddef.h>
+#include <stdlib.h>
+#endif
+#include <sys/types.h>
+
+#if defined(_WIN64)
+typedef signed __int64 ssize_t;
+#else
+typedef signed int ssize_t;
+#endif /* _WIN64 */
+
+#endif /* _MSC_VER */
+
+/* Set defaults for constants which we cannot find. */
+#if !defined(INT64_MAX)
+# define INT64_MAX 9223372036854775807LL
+#endif
+#if !defined(INT64_MIN)
+# define INT64_MIN ((-INT64_MAX)-1)
+#endif
+#if !defined(UINT64_MAX)
+# define UINT64_MAX 0xffffffffffffffffULL
+#endif
+
+#ifndef HUGE_VALF
+#define HUGE_VALF (float)HUGE_VAL
+#endif
+
+#endif /* LLVM_C_DATATYPES_H */
diff --git a/third_party/llvm-project/include/llvm-c/DisassemblerTypes.h b/third_party/llvm-project/include/llvm-c/DisassemblerTypes.h
new file mode 100644
index 000000000..389e5ee45
--- /dev/null
+++ b/third_party/llvm-project/include/llvm-c/DisassemblerTypes.h
@@ -0,0 +1,160 @@
+/*===-- llvm-c/DisassemblerTypedefs.h -----------------------------*- C -*-===*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*/
+
+#ifndef LLVM_DISASSEMBLER_TYPES_H
+#define LLVM_DISASSEMBLER_TYPES_H
+
+#include "llvm-c/DataTypes.h"
+#ifdef __cplusplus
+#include <cstddef>
+#else
+#include <stddef.h>
+#endif
+
+/**
+ * An opaque reference to a disassembler context.
+ */
+typedef void *LLVMDisasmContextRef;
+
+/**
+ * The type for the operand information call back function.  This is called to
+ * get the symbolic information for an operand of an instruction.  Typically
+ * this is from the relocation information, symbol table, etc.  That block of
+ * information is saved when the disassembler context is created and passed to
+ * the call back in the DisInfo parameter.  The instruction containing operand
+ * is at the PC parameter.  For some instruction sets, there can be more than
+ * one operand with symbolic information.  To determine the symbolic operand
+ * information for each operand, the bytes for the specific operand in the
+ * instruction are specified by the Offset parameter and its byte widith is the
+ * size parameter.  For instructions sets with fixed widths and one symbolic
+ * operand per instruction, the Offset parameter will be zero and Size parameter
+ * will be the instruction width.  The information is returned in TagBuf and is
+ * Triple specific with its specific information defined by the value of
+ * TagType for that Triple.  If symbolic information is returned the function
+ * returns 1, otherwise it returns 0.
+ */
+typedef int (*LLVMOpInfoCallback)(void *DisInfo, uint64_t PC,
+                                  uint64_t Offset, uint64_t Size,
+                                  int TagType, void *TagBuf);
+
+/**
+ * The initial support in LLVM MC for the most general form of a relocatable
+ * expression is "AddSymbol - SubtractSymbol + Offset".  For some Darwin targets
+ * this full form is encoded in the relocation information so that AddSymbol and
+ * SubtractSymbol can be link edited independent of each other.  Many other
+ * platforms only allow a relocatable expression of the form AddSymbol + Offset
+ * to be encoded.
+ *
+ * The LLVMOpInfoCallback() for the TagType value of 1 uses the struct
+ * LLVMOpInfo1.  The value of the relocatable expression for the operand,
+ * including any PC adjustment, is passed in to the call back in the Value
+ * field.  The symbolic information about the operand is returned using all
+ * the fields of the structure with the Offset of the relocatable expression
+ * returned in the Value field.  It is possible that some symbols in the
+ * relocatable expression were assembly temporary symbols, for example
+ * "Ldata - LpicBase + constant", and only the Values of the symbols without
+ * symbol names are present in the relocation information.  The VariantKind
+ * type is one of the Target specific #defines below and is used to print
+ * operands like "_foo@GOT", ":lower16:_foo", etc.
+ */
+struct LLVMOpInfoSymbol1 {
+  uint64_t Present;  /* 1 if this symbol is present */
+  const char *Name;  /* symbol name if not NULL */
+  uint64_t Value;    /* symbol value if name is NULL */
+};
+
+struct LLVMOpInfo1 {
+  struct LLVMOpInfoSymbol1 AddSymbol;
+  struct LLVMOpInfoSymbol1 SubtractSymbol;
+  uint64_t Value;
+  uint64_t VariantKind;
+};
+
+/**
+ * The operand VariantKinds for symbolic disassembly.
+ */
+#define LLVMDisassembler_VariantKind_None 0 /* all targets */
+
+/**
+ * The ARM target VariantKinds.
+ */
+#define LLVMDisassembler_VariantKind_ARM_HI16 1 /* :upper16: */
+#define LLVMDisassembler_VariantKind_ARM_LO16 2 /* :lower16: */
+
+/**
+ * The ARM64 target VariantKinds.
+ */
+#define LLVMDisassembler_VariantKind_ARM64_PAGE       1 /* @page */
+#define LLVMDisassembler_VariantKind_ARM64_PAGEOFF    2 /* @pageoff */
+#define LLVMDisassembler_VariantKind_ARM64_GOTPAGE    3 /* @gotpage */
+#define LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF 4 /* @gotpageoff */
+#define LLVMDisassembler_VariantKind_ARM64_TLVP       5 /* @tvlppage */
+#define LLVMDisassembler_VariantKind_ARM64_TLVOFF     6 /* @tvlppageoff */
+
+/**
+ * The type for the symbol lookup function.  This may be called by the
+ * disassembler for things like adding a comment for a PC plus a constant
+ * offset load instruction to use a symbol name instead of a load address value.
+ * It is passed the block information is saved when the disassembler context is
+ * created and the ReferenceValue to look up as a symbol.  If no symbol is found
+ * for the ReferenceValue NULL is returned.  The ReferenceType of the
+ * instruction is passed indirectly as is the PC of the instruction in
+ * ReferencePC.  If the output reference can be determined its type is returned
+ * indirectly in ReferenceType along with ReferenceName if any, or that is set
+ * to NULL.
+ */
+typedef const char *(*LLVMSymbolLookupCallback)(void *DisInfo,
+                                                uint64_t ReferenceValue,
+                                                uint64_t *ReferenceType,
+                                                uint64_t ReferencePC,
+                                                const char **ReferenceName);
+/**
+ * The reference types on input and output.
+ */
+/* No input reference type or no output reference type. */
+#define LLVMDisassembler_ReferenceType_InOut_None 0
+
+/* The input reference is from a branch instruction. */
+#define LLVMDisassembler_ReferenceType_In_Branch 1
+/* The input reference is from a PC relative load instruction. */
+#define LLVMDisassembler_ReferenceType_In_PCrel_Load 2
+
+/* The input reference is from an ARM64::ADRP instruction. */
+#define LLVMDisassembler_ReferenceType_In_ARM64_ADRP 0x100000001
+/* The input reference is from an ARM64::ADDXri instruction. */
+#define LLVMDisassembler_ReferenceType_In_ARM64_ADDXri 0x100000002
+/* The input reference is from an ARM64::LDRXui instruction. */
+#define LLVMDisassembler_ReferenceType_In_ARM64_LDRXui 0x100000003
+/* The input reference is from an ARM64::LDRXl instruction. */
+#define LLVMDisassembler_ReferenceType_In_ARM64_LDRXl 0x100000004
+/* The input reference is from an ARM64::ADR instruction. */
+#define LLVMDisassembler_ReferenceType_In_ARM64_ADR 0x100000005
+
+/* The output reference is to as symbol stub. */
+#define LLVMDisassembler_ReferenceType_Out_SymbolStub 1
+/* The output reference is to a symbol address in a literal pool. */
+#define LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr 2
+/* The output reference is to a cstring address in a literal pool. */
+#define LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr 3
+
+/* The output reference is to a Objective-C CoreFoundation string. */
+#define LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref 4
+/* The output reference is to a Objective-C message. */
+#define LLVMDisassembler_ReferenceType_Out_Objc_Message 5
+/* The output reference is to a Objective-C message ref. */
+#define LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref 6
+/* The output reference is to a Objective-C selector ref. */
+#define LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref 7
+/* The output reference is to a Objective-C class ref. */
+#define LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref 8
+
+/* The output reference is to a C++ symbol name. */
+#define LLVMDisassembler_ReferenceType_DeMangled_Name 9
+
+#endif
diff --git a/third_party/llvm-project/include/llvm-c/Error.h b/third_party/llvm-project/include/llvm-c/Error.h
new file mode 100644
index 000000000..52943063c
--- /dev/null
+++ b/third_party/llvm-project/include/llvm-c/Error.h
@@ -0,0 +1,69 @@
+/*===------- llvm-c/Error.h - llvm::Error class C Interface -------*- C -*-===*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*|
+|*                                                                            *|
+|* This file defines the C interface to LLVM's Error class.                   *|
+|*                                                                            *|
+\*===----------------------------------------------------------------------===*/
+
+#ifndef LLVM_C_ERROR_H
+#define LLVM_C_ERROR_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define LLVMErrorSuccess 0
+
+/**
+ * Opaque reference to an error instance. Null serves as the 'success' value.
+ */
+typedef struct LLVMOpaqueError *LLVMErrorRef;
+
+/**
+ * Error type identifier.
+ */
+typedef const void *LLVMErrorTypeId;
+
+/**
+ * Returns the type id for the given error instance, which must be a failure
+ * value (i.e. non-null).
+ */
+LLVMErrorTypeId LLVMGetErrorTypeId(LLVMErrorRef Err);
+
+/**
+ * Dispose of the given error without handling it. This operation consumes the
+ * error, and the given LLVMErrorRef value is not usable once this call returns.
+ * Note: This method *only* needs to be called if the error is not being passed
+ * to some other consuming operation, e.g. LLVMGetErrorMessage.
+ */
+void LLVMConsumeError(LLVMErrorRef Err);
+
+/**
+ * Returns the given string's error message. This operation consumes the error,
+ * and the given LLVMErrorRef value is not usable once this call returns.
+ * The caller is responsible for disposing of the string by calling
+ * LLVMDisposeErrorMessage.
+ */
+char *LLVMGetErrorMessage(LLVMErrorRef Err);
+
+/**
+ * Dispose of the given error message.
+ */
+void LLVMDisposeErrorMessage(char *ErrMsg);
+
+/**
+ * Returns the type id for llvm StringError.
+ */
+LLVMErrorTypeId LLVMGetStringErrorTypeId(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/third_party/llvm-project/include/llvm-c/ErrorHandling.h b/third_party/llvm-project/include/llvm-c/ErrorHandling.h
new file mode 100644
index 000000000..4927349d8
--- /dev/null
+++ b/third_party/llvm-project/include/llvm-c/ErrorHandling.h
@@ -0,0 +1,49 @@
+/*===-- llvm-c/ErrorHandling.h - Error Handling C Interface -------*- C -*-===*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*|
+|*                                                                            *|
+|* This file defines the C interface to LLVM's error handling mechanism.      *|
+|*                                                                            *|
+\*===----------------------------------------------------------------------===*/
+
+#ifndef LLVM_C_ERROR_HANDLING_H
+#define LLVM_C_ERROR_HANDLING_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef void (*LLVMFatalErrorHandler)(const char *Reason);
+
+/**
+ * Install a fatal error handler. By default, if LLVM detects a fatal error, it
+ * will call exit(1). This may not be appropriate in many contexts. For example,
+ * doing exit(1) will bypass many crash reporting/tracing system tools. This
+ * function allows you to install a callback that will be invoked prior to the
+ * call to exit(1).
+ */
+void LLVMInstallFatalErrorHandler(LLVMFatalErrorHandler Handler);
+
+/**
+ * Reset the fatal error handler. This resets LLVM's fatal error handling
+ * behavior to the default.
+ */
+void LLVMResetFatalErrorHandler(void);
+
+/**
+ * Enable LLVM's built-in stack trace code. This intercepts the OS's crash
+ * signals and prints which component of LLVM you were in at the time if the
+ * crash.
+ */
+void LLVMEnablePrettyStackTrace(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/third_party/llvm-project/include/llvm-c/Types.h b/third_party/llvm-project/include/llvm-c/Types.h
new file mode 100644
index 000000000..612c7d3ef
--- /dev/null
+++ b/third_party/llvm-project/include/llvm-c/Types.h
@@ -0,0 +1,179 @@
+/*===-- llvm-c/Support.h - C Interface Types declarations ---------*- C -*-===*\
+|*                                                                            *|
+|* Part of the LLVM Project, under the Apache License v2.0 with LLVM          *|
+|* Exceptions.                                                                *|
+|* See https://llvm.org/LICENSE.txt for license information.                  *|
+|* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception                    *|
+|*                                                                            *|
+|*===----------------------------------------------------------------------===*|
+|*                                                                            *|
+|* This file defines types used by the C interface to LLVM.                   *|
+|*                                                                            *|
+\*===----------------------------------------------------------------------===*/
+
+#ifndef LLVM_C_TYPES_H
+#define LLVM_C_TYPES_H
+
+#include "llvm-c/DataTypes.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @defgroup LLVMCSupportTypes Types and Enumerations
+ *
+ * @{
+ */
+
+typedef int LLVMBool;
+
+/* Opaque types. */
+
+/**
+ * LLVM uses a polymorphic type hierarchy which C cannot represent, therefore
+ * parameters must be passed as base types. Despite the declared types, most
+ * of the functions provided operate only on branches of the type hierarchy.
+ * The declared parameter names are descriptive and specify which type is
+ * required. Additionally, each type hierarchy is documented along with the
+ * functions that operate upon it. For more detail, refer to LLVM's C++ code.
+ * If in doubt, refer to Core.cpp, which performs parameter downcasts in the
+ * form unwrap<RequiredType>(Param).
+ */
+
+/**
+ * Used to pass regions of memory through LLVM interfaces.
+ *
+ * @see llvm::MemoryBuffer
+ */
+typedef struct LLVMOpaqueMemoryBuffer *LLVMMemoryBufferRef;
+
+/**
+ * The top-level container for all LLVM global data. See the LLVMContext class.
+ */
+typedef struct LLVMOpaqueContext *LLVMContextRef;
+
+/**
+ * The top-level container for all other LLVM Intermediate Representation (IR)
+ * objects.
+ *
+ * @see llvm::Module
+ */
+typedef struct LLVMOpaqueModule *LLVMModuleRef;
+
+/**
+ * Each value in the LLVM IR has a type, an LLVMTypeRef.
+ *
+ * @see llvm::Type
+ */
+typedef struct LLVMOpaqueType *LLVMTypeRef;
+
+/**
+ * Represents an individual value in LLVM IR.
+ *
+ * This models llvm::Value.
+ */
+typedef struct LLVMOpaqueValue *LLVMValueRef;
+
+/**
+ * Represents a basic block of instructions in LLVM IR.
+ *
+ * This models llvm::BasicBlock.
+ */
+typedef struct LLVMOpaqueBasicBlock *LLVMBasicBlockRef;
+
+/**
+ * Represents an LLVM Metadata.
+ *
+ * This models llvm::Metadata.
+ */
+typedef struct LLVMOpaqueMetadata *LLVMMetadataRef;
+
+/**
+ * Represents an LLVM Named Metadata Node.
+ *
+ * This models llvm::NamedMDNode.
+ */
+typedef struct LLVMOpaqueNamedMDNode *LLVMNamedMDNodeRef;
+
+/**
+ * Represents an entry in a Global Object's metadata attachments.
+ *
+ * This models std::pair<unsigned, MDNode *>
+ */
+typedef struct LLVMOpaqueValueMetadataEntry LLVMValueMetadataEntry;
+
+/**
+ * Represents an LLVM basic block builder.
+ *
+ * This models llvm::IRBuilder.
+ */
+typedef struct LLVMOpaqueBuilder *LLVMBuilderRef;
+
+/**
+ * Represents an LLVM debug info builder.
+ *
+ * This models llvm::DIBuilder.
+ */
+typedef struct LLVMOpaqueDIBuilder *LLVMDIBuilderRef;
+
+/**
+ * Interface used to provide a module to JIT or interpreter.
+ * This is now just a synonym for llvm::Module, but we have to keep using the
+ * different type to keep binary compatibility.
+ */
+typedef struct LLVMOpaqueModuleProvider *LLVMModuleProviderRef;
+
+/** @see llvm::PassManagerBase */
+typedef struct LLVMOpaquePassManager *LLVMPassManagerRef;
+
+/** @see llvm::PassRegistry */
+typedef struct LLVMOpaquePassRegistry *LLVMPassRegistryRef;
+
+/**
+ * Used to get the users and usees of a Value.
+ *
+ * @see llvm::Use */
+typedef struct LLVMOpaqueUse *LLVMUseRef;
+
+/**
+ * Used to represent an attributes.
+ *
+ * @see llvm::Attribute
+ */
+typedef struct LLVMOpaqueAttributeRef *LLVMAttributeRef;
+
+/**
+ * @see llvm::DiagnosticInfo
+ */
+typedef struct LLVMOpaqueDiagnosticInfo *LLVMDiagnosticInfoRef;
+
+/**
+ * @see llvm::Comdat
+ */
+typedef struct LLVMComdat *LLVMComdatRef;
+
+/**
+ * @see llvm::Module::ModuleFlagEntry
+ */
+typedef struct LLVMOpaqueModuleFlagEntry LLVMModuleFlagEntry;
+
+/**
+ * @see llvm::JITEventListener
+ */
+typedef struct LLVMOpaqueJITEventListener *LLVMJITEventListenerRef;
+
+/**
+ * @see llvm::object::Binary
+ */
+typedef struct LLVMOpaqueBinary *LLVMBinaryRef;
+
+/**
+ * @}
+ */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif