Ekdohibs
/
zig
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/master' into llvm6

master
Andrew Kelley 2017-11-29 16:34:50 -05:00
parent 9a4da6c8d8 7066283004
commit 91ef68f9b1
30 changed files with 3571 additions and 1277 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
CMakeLists.txt
View File

@ -339,7 +339,7 @@ set(ZIG_SOURCES
"${CMAKE_SOURCE_DIR}/src/target.cpp"
"${CMAKE_SOURCE_DIR}/src/tokenizer.cpp"
"${CMAKE_SOURCE_DIR}/src/util.cpp"
"${CMAKE_SOURCE_DIR}/src/parsec.cpp"
"${CMAKE_SOURCE_DIR}/src/translate_c.cpp"
"${CMAKE_SOURCE_DIR}/src/zig_llvm.cpp"
)

2
build.zig
View File

@ -58,5 +58,5 @@ pub fn build(b: &Builder) {
test_step.dependOn(tests.addCompileErrorTests(b, test_filter));
test_step.dependOn(tests.addAssembleAndLinkTests(b, test_filter));
test_step.dependOn(tests.addDebugSafetyTests(b, test_filter));
test_step.dependOn(tests.addParseCTests(b, test_filter));
test_step.dependOn(tests.addTranslateCTests(b, test_filter));
}

2
ci/travis_osx_script
View File

@ -22,4 +22,4 @@ make install
./zig build --build-file ../build.zig test-compile-errors --verbose
./zig build --build-file ../build.zig test-asm-link --verbose
./zig build --build-file ../build.zig test-debug-safety --verbose
./zig build --build-file ../build.zig test-parsec --verbose
./zig build --build-file ../build.zig test-translate-c --verbose

8
doc/home.html.in
View File

@ -70,10 +70,14 @@
<li><a href="#mersenne">Mersenne Twister Random Number Generator</a></li>
</ul>
<h3 id="hello">Hello World</h3>
<pre><code class="zig">const io = @import("std").io;
<pre><code class="zig">const std = @import("std");
pub fn main() -&gt; %void {
%return io.stdout.printf("Hello, world!\n");
// If this program is run without stdout attached, exit with an error.
var stdout_file = %return std.io.getStdOut();
// If this program encounters pipe failure when printing to stdout, exit
// with an error.
%return stdout_file.write("Hello, world!\n");
}</code></pre>
<p>Build this with:</p>
<pre>zig build-exe hello.zig</pre>

19
doc/langref.html.in
View File

@ -75,6 +75,7 @@
<li><a href="#slices">Slices</a></li>
<li><a href="#struct">struct</a></li>
<li><a href="#enum">enum</a></li>
<li><a href="#union">union</a></li>
<li><a href="#switch">switch</a></li>
<li><a href="#while">while</a></li>
<li><a href="#for">for</a></li>
@ -209,6 +210,7 @@
<li><a href="#undef-invalid-error-code">Invalid Error Code</a></li>
<li><a href="#undef-invalid-enum-cast">Invalid Enum Cast</a></li>
<li><a href="#undef-incorrect-pointer-alignment">Incorrect Pointer Alignment</a></li>
<li><a href="#undef-bad-union-field">Wrong Union Field Access</a></li>
</ul>
</li>
<li><a href="#memory">Memory</a></li>
@ -2189,6 +2191,8 @@ Test 4/4 enum builtins...OK</code></pre>
<li><a href="#builtin-enumTagName">@enumTagName</a></li>
<li><a href="#builtin-memberCount">@memberCount</a></li>
</ul>
<h2 id="union">union</h2>
<p>TODO union documentation</p>
<h2 id="switch">switch</h2>
<pre><code class="zig">const assert = @import("std").debug.assert;
const builtin = @import("builtin");
@ -5117,6 +5121,9 @@ comptime {
<h3 id="undef-incorrect-pointer-alignment">Incorrect Pointer Alignment</h3>
<p>TODO</p>
<h3 id="undef-bad-union-field">Wrong Union Field Access</h3>
<p>TODO</p>
<h2 id="memory">Memory</h2>
<p>TODO: explain no default allocator in zig</p>
<p>TODO: show how to use the allocator interface</p>
@ -5405,10 +5412,14 @@ const c = @cImport({
export fn decode_base_64(dest_ptr: &amp;u8, dest_len: usize,
source_ptr: &amp;const u8, source_len: usize) -&gt; usize
{
const src = source_ptr[0...source_len];
const dest = dest_ptr[0...dest_len];
return base64.decode(dest, src).len;
}</code></pre>
const src = source_ptr[0..source_len];
const dest = dest_ptr[0..dest_len];
const base64_decoder = base64.standard_decoder_unsafe;
const decoded_size = base64_decoder.calcSize(src);
base64_decoder.decode(dest[0..decoded_size], src);
return decoded_size;
}
</code></pre>
<h4>test.c</h4>
<pre><code class="c">// This header is generated by zig from base64.zig
#include "base64.h"

5
example/mix_o_files/base64.zig
View File

@ -3,5 +3,8 @@ const base64 = @import("std").base64;
export fn decode_base_64(dest_ptr: &u8, dest_len: usize, source_ptr: &const u8, source_len: usize) -> usize {
const src = source_ptr[0..source_len];
const dest = dest_ptr[0..dest_len];
return base64.decode(dest, src).len;
const base64_decoder = base64.standard_decoder_unsafe;
const decoded_size = base64_decoder.calcSize(src);
base64_decoder.decode(dest[0..decoded_size], src);
return decoded_size;
}

2
src-self-hosted/main.zig
View File

@ -208,7 +208,7 @@ fn printUsage(outstream: &io.OutStream) -> %void {
\\ build-exe [source] create executable from source or object files
\\ build-lib [source] create library from source or object files
\\ build-obj [source] create object from source or assembly
\\ parsec [source] convert c code to zig code
\\ translate-c [source] convert c code to zig code
\\ targets list available compilation targets
\\ test [source] create and run a test build
\\ version print version number and exit

57
src/all_types.hpp
View File

@ -36,6 +36,7 @@ struct IrInstructionCast;
struct IrBasicBlock;
struct ScopeDecls;
struct ZigWindowsSDK;
struct Tld;
struct IrGotoItem {
AstNode *source_node;
@ -59,7 +60,9 @@ struct IrExecutable {
Buf *c_import_buf;
AstNode *source_node;
IrExecutable *parent_exec;
IrExecutable *source_exec;
Scope *begin_scope;
ZigList<Tld *> tld_list;
};
enum OutType {
@ -73,6 +76,7 @@ enum ConstParentId {
ConstParentIdNone,
ConstParentIdStruct,
ConstParentIdArray,
ConstParentIdUnion,
};
struct ConstParent {
@ -87,6 +91,9 @@ struct ConstParent {
ConstExprValue *struct_val;
size_t field_index;
} p_struct;
struct {
ConstExprValue *union_val;
} p_union;
} data;
};
@ -100,6 +107,12 @@ struct ConstStructValue {
ConstParent parent;
};
struct ConstUnionValue {
uint64_t tag;
ConstExprValue *payload;
ConstParent parent;
};
enum ConstArraySpecial {
ConstArraySpecialNone,
ConstArraySpecialUndef,
@ -238,6 +251,7 @@ struct ConstExprValue {
ErrorTableEntry *x_pure_err;
ConstEnumValue x_enum;
ConstStructValue x_struct;
ConstUnionValue x_union;
ConstArrayValue x_array;
ConstPtrValue x_ptr;
ImportTableEntry *x_import;
@ -336,6 +350,13 @@ struct TypeEnumField {
uint32_t gen_index;
};
struct TypeUnionField {
Buf *name;
TypeTableEntry *type_entry;
uint32_t value;
uint32_t gen_index;
};
enum NodeType {
NodeTypeRoot,
NodeTypeFnProto,
@ -1021,14 +1042,19 @@ struct TypeTableEntryEnumTag {
LLVMValueRef name_table;
};
uint32_t type_ptr_hash(const TypeTableEntry *ptr);
bool type_ptr_eql(const TypeTableEntry *a, const TypeTableEntry *b);
struct TypeTableEntryUnion {
AstNode *decl_node;
ContainerLayout layout;
uint32_t src_field_count;
uint32_t gen_field_count;
TypeStructField *fields;
uint64_t size_bytes;
TypeUnionField *fields;
bool is_invalid; // true if any fields are invalid
TypeTableEntry *tag_type;
LLVMTypeRef union_type_ref;
ScopeDecls *decls_scope;
// set this flag temporarily to detect infinite loops
@ -1039,6 +1065,13 @@ struct TypeTableEntryUnion {
bool zero_bits_loop_flag;
bool zero_bits_known;
uint32_t abi_alignment; // also figured out with zero_bits pass
size_t gen_union_index;
size_t gen_tag_index;
uint32_t union_size_bytes;
TypeTableEntry *most_aligned_union_member;
};
struct FnGenParamInfo {
@ -1287,6 +1320,7 @@ enum PanicMsgId {
PanicMsgIdUnwrapMaybeFail,
PanicMsgIdInvalidErrorCode,
PanicMsgIdIncorrectAlignment,
PanicMsgIdBadUnionField,
PanicMsgIdCount,
};
@ -1796,6 +1830,7 @@ enum IrInstructionId {
IrInstructionIdFieldPtr,
IrInstructionIdStructFieldPtr,
IrInstructionIdEnumFieldPtr,
IrInstructionIdUnionFieldPtr,
IrInstructionIdElemPtr,
IrInstructionIdVarPtr,
IrInstructionIdCall,
@ -1805,6 +1840,7 @@ enum IrInstructionId {
IrInstructionIdContainerInitList,
IrInstructionIdContainerInitFields,
IrInstructionIdStructInit,
IrInstructionIdUnionInit,
IrInstructionIdUnreachable,
IrInstructionIdTypeOf,
IrInstructionIdToPtrType,
@ -2060,6 +2096,14 @@ struct IrInstructionEnumFieldPtr {
bool is_const;
};
struct IrInstructionUnionFieldPtr {
IrInstruction base;
IrInstruction *union_ptr;
TypeUnionField *field;
bool is_const;
};
struct IrInstructionElemPtr {
IrInstruction base;
@ -2150,6 +2194,15 @@ struct IrInstructionStructInit {
LLVMValueRef tmp_ptr;
};
struct IrInstructionUnionInit {
IrInstruction base;
TypeTableEntry *union_type;
TypeUnionField *field;
IrInstruction *init_value;
LLVMValueRef tmp_ptr;
};
struct IrInstructionUnreachable {
IrInstruction base;
};

343
src/analyze.cpp
View File

@ -28,7 +28,7 @@ static void resolve_union_zero_bits(CodeGen *g, TypeTableEntry *union_type);
ErrorMsg *add_node_error(CodeGen *g, AstNode *node, Buf *msg) {
if (node->owner->c_import_node != nullptr) {
// if this happens, then parsec generated code that
// if this happens, then translate_c generated code that
// failed semantic analysis, which isn't supposed to happen
ErrorMsg *err = add_node_error(g, node->owner->c_import_node,
buf_sprintf("compiler bug: @cImport generated invalid zig code"));
@ -48,7 +48,7 @@ ErrorMsg *add_node_error(CodeGen *g, AstNode *node, Buf *msg) {
ErrorMsg *add_error_note(CodeGen *g, ErrorMsg *parent_msg, AstNode *node, Buf *msg) {
if (node->owner->c_import_node != nullptr) {
// if this happens, then parsec generated code that
// if this happens, then translate_c generated code that
// failed semantic analysis, which isn't supposed to happen
Buf *note_path = buf_create_from_str("?.c");
@ -338,7 +338,7 @@ TypeTableEntry *get_smallest_unsigned_int_type(CodeGen *g, uint64_t x) {
TypeTableEntry *get_pointer_to_type_extra(CodeGen *g, TypeTableEntry *child_type, bool is_const,
bool is_volatile, uint32_t byte_alignment, uint32_t bit_offset, uint32_t unaligned_bit_count)
{
assert(child_type->id != TypeTableEntryIdInvalid);
assert(!type_is_invalid(child_type));
TypeId type_id = {};
TypeTableEntry **parent_pointer = nullptr;
@ -1008,11 +1008,12 @@ TypeTableEntry *get_partial_container_type(CodeGen *g, Scope *scope, ContainerKi
}
size_t line = decl_node ? decl_node->line : 0;
unsigned dwarf_kind = ZigLLVMTag_DW_structure_type();
ImportTableEntry *import = get_scope_import(scope);
entry->type_ref = LLVMStructCreateNamed(LLVMGetGlobalContext(), name);
entry->di_type = ZigLLVMCreateReplaceableCompositeType(g->dbuilder,
ZigLLVMTag_DW_structure_type(), name,
dwarf_kind, name,
ZigLLVMFileToScope(import->di_file), import->di_file, (unsigned)(line + 1));
buf_init_from_str(&entry->name, name);
@ -1285,7 +1286,7 @@ static void resolve_enum_type(CodeGen *g, TypeTableEntry *enum_type) {
return;
resolve_enum_zero_bits(g, enum_type);
if (enum_type->data.enumeration.is_invalid)
if (type_is_invalid(enum_type))
return;
AstNode *decl_node = enum_type->data.enumeration.decl_node;
@ -1834,7 +1835,246 @@ static void resolve_struct_type(CodeGen *g, TypeTableEntry *struct_type) {
}
static void resolve_union_type(CodeGen *g, TypeTableEntry *union_type) {
zig_panic("TODO");
assert(union_type->id == TypeTableEntryIdUnion);
if (union_type->data.unionation.complete)
return;
resolve_union_zero_bits(g, union_type);
if (type_is_invalid(union_type))
return;
AstNode *decl_node = union_type->data.unionation.decl_node;
if (union_type->data.unionation.embedded_in_current) {
if (!union_type->data.unionation.reported_infinite_err) {
union_type->data.unionation.reported_infinite_err = true;
add_node_error(g, decl_node, buf_sprintf("union '%s' contains itself", buf_ptr(&union_type->name)));
}
return;
}
assert(!union_type->data.unionation.zero_bits_loop_flag);
assert(decl_node->type == NodeTypeContainerDecl);
assert(union_type->di_type);
uint32_t field_count = union_type->data.unionation.src_field_count;
assert(union_type->data.unionation.fields);
uint32_t gen_field_count = union_type->data.unionation.gen_field_count;
ZigLLVMDIType **union_inner_di_types = allocate<ZigLLVMDIType*>(gen_field_count);
TypeTableEntry *most_aligned_union_member = nullptr;
uint64_t size_of_most_aligned_member_in_bits = 0;
uint64_t biggest_align_in_bits = 0;
uint64_t biggest_size_in_bits = 0;
bool auto_layout = (union_type->data.unionation.layout == ContainerLayoutAuto);
ZigLLVMDIEnumerator **di_enumerators = allocate<ZigLLVMDIEnumerator*>(field_count);
Scope *scope = &union_type->data.unionation.decls_scope->base;
ImportTableEntry *import = get_scope_import(scope);
// set temporary flag
union_type->data.unionation.embedded_in_current = true;
for (uint32_t i = 0; i < field_count; i += 1) {
AstNode *field_node = decl_node->data.container_decl.fields.at(i);
TypeUnionField *type_union_field = &union_type->data.unionation.fields[i];
TypeTableEntry *field_type = type_union_field->type_entry;
ensure_complete_type(g, field_type);
if (type_is_invalid(field_type)) {
union_type->data.unionation.is_invalid = true;
continue;
}
if (!type_has_bits(field_type))
continue;
di_enumerators[i] = ZigLLVMCreateDebugEnumerator(g->dbuilder, buf_ptr(type_union_field->name), i);
uint64_t store_size_in_bits = 8*LLVMStoreSizeOfType(g->target_data_ref, field_type->type_ref);
uint64_t abi_align_in_bits = 8*LLVMABIAlignmentOfType(g->target_data_ref, field_type->type_ref);
assert(store_size_in_bits > 0);
assert(abi_align_in_bits > 0);
union_inner_di_types[type_union_field->gen_index] = ZigLLVMCreateDebugMemberType(g->dbuilder,
ZigLLVMTypeToScope(union_type->di_type), buf_ptr(type_union_field->name),
import->di_file, (unsigned)(field_node->line + 1),
store_size_in_bits,
abi_align_in_bits,
0,
0, field_type->di_type);
biggest_size_in_bits = max(biggest_size_in_bits, store_size_in_bits);
if (!most_aligned_union_member || abi_align_in_bits > biggest_align_in_bits) {
most_aligned_union_member = field_type;
biggest_align_in_bits = abi_align_in_bits;
size_of_most_aligned_member_in_bits = store_size_in_bits;
}
}
// unset temporary flag
union_type->data.unionation.embedded_in_current = false;
union_type->data.unionation.complete = true;
union_type->data.unionation.union_size_bytes = biggest_size_in_bits / 8;
union_type->data.unionation.most_aligned_union_member = most_aligned_union_member;
if (union_type->data.unionation.is_invalid)
return;
if (union_type->zero_bits) {
union_type->type_ref = LLVMVoidType();
uint64_t debug_size_in_bits = 0;
uint64_t debug_align_in_bits = 0;
ZigLLVMDIType **di_root_members = nullptr;
size_t debug_member_count = 0;
ZigLLVMDIType *replacement_di_type = ZigLLVMCreateDebugUnionType(g->dbuilder,
ZigLLVMFileToScope(import->di_file),
buf_ptr(&union_type->name),
import->di_file, (unsigned)(decl_node->line + 1),
debug_size_in_bits,
debug_align_in_bits,
0, di_root_members, (int)debug_member_count, 0, "");
ZigLLVMReplaceTemporary(g->dbuilder, union_type->di_type, replacement_di_type);
union_type->di_type = replacement_di_type;
return;
}
assert(most_aligned_union_member != nullptr);
bool want_safety = auto_layout && (field_count >= 2);
uint64_t padding_in_bits = biggest_size_in_bits - size_of_most_aligned_member_in_bits;
if (!want_safety) {
if (padding_in_bits > 0) {
TypeTableEntry *u8_type = get_int_type(g, false, 8);
TypeTableEntry *padding_array = get_array_type(g, u8_type, padding_in_bits / 8);
LLVMTypeRef union_element_types[] = {
most_aligned_union_member->type_ref,
padding_array->type_ref,
};
LLVMStructSetBody(union_type->type_ref, union_element_types, 2, false);
} else {
LLVMStructSetBody(union_type->type_ref, &most_aligned_union_member->type_ref, 1, false);
}
union_type->data.unionation.union_type_ref = union_type->type_ref;
union_type->data.unionation.gen_tag_index = SIZE_MAX;
union_type->data.unionation.gen_union_index = SIZE_MAX;
assert(8*LLVMABIAlignmentOfType(g->target_data_ref, union_type->type_ref) >= biggest_align_in_bits);
assert(8*LLVMStoreSizeOfType(g->target_data_ref, union_type->type_ref) >= biggest_size_in_bits);
// create debug type for union
ZigLLVMDIType *replacement_di_type = ZigLLVMCreateDebugUnionType(g->dbuilder,
ZigLLVMFileToScope(import->di_file), buf_ptr(&union_type->name),
import->di_file, (unsigned)(decl_node->line + 1),
biggest_size_in_bits, biggest_align_in_bits, 0, union_inner_di_types,
gen_field_count, 0, "");
ZigLLVMReplaceTemporary(g->dbuilder, union_type->di_type, replacement_di_type);
union_type->di_type = replacement_di_type;
return;
}
LLVMTypeRef union_type_ref;
if (padding_in_bits > 0) {
TypeTableEntry *u8_type = get_int_type(g, false, 8);
TypeTableEntry *padding_array = get_array_type(g, u8_type, padding_in_bits / 8);
LLVMTypeRef union_element_types[] = {
most_aligned_union_member->type_ref,
padding_array->type_ref,
};
union_type_ref = LLVMStructType(union_element_types, 2, false);
} else {
union_type_ref = most_aligned_union_member->type_ref;
}
union_type->data.unionation.union_type_ref = union_type_ref;
assert(8*LLVMABIAlignmentOfType(g->target_data_ref, union_type_ref) >= biggest_align_in_bits);
assert(8*LLVMStoreSizeOfType(g->target_data_ref, union_type_ref) >= biggest_size_in_bits);
// create llvm type for root struct
TypeTableEntry *tag_int_type = get_smallest_unsigned_int_type(g, field_count - 1);
TypeTableEntry *tag_type_entry = tag_int_type;
union_type->data.unionation.tag_type = tag_type_entry;
uint64_t align_of_tag_in_bits = 8*LLVMABIAlignmentOfType(g->target_data_ref, tag_int_type->type_ref);
if (align_of_tag_in_bits >= biggest_align_in_bits) {
union_type->data.unionation.gen_tag_index = 0;
union_type->data.unionation.gen_union_index = 1;
} else {
union_type->data.unionation.gen_union_index = 0;
union_type->data.unionation.gen_tag_index = 1;
}
LLVMTypeRef root_struct_element_types[2];
root_struct_element_types[union_type->data.unionation.gen_tag_index] = tag_type_entry->type_ref;
root_struct_element_types[union_type->data.unionation.gen_union_index] = union_type_ref;
LLVMStructSetBody(union_type->type_ref, root_struct_element_types, 2, false);
// create debug type for root struct
// create debug type for tag
uint64_t tag_debug_size_in_bits = 8*LLVMStoreSizeOfType(g->target_data_ref, tag_type_entry->type_ref);
uint64_t tag_debug_align_in_bits = 8*LLVMABIAlignmentOfType(g->target_data_ref, tag_type_entry->type_ref);
ZigLLVMDIType *tag_di_type = ZigLLVMCreateDebugEnumerationType(g->dbuilder,
ZigLLVMTypeToScope(union_type->di_type), "AnonEnum",
import->di_file, (unsigned)(decl_node->line + 1),
tag_debug_size_in_bits, tag_debug_align_in_bits, di_enumerators, field_count,
tag_type_entry->di_type, "");
// create debug type for union
ZigLLVMDIType *union_di_type = ZigLLVMCreateDebugUnionType(g->dbuilder,
ZigLLVMTypeToScope(union_type->di_type), "AnonUnion",
import->di_file, (unsigned)(decl_node->line + 1),
biggest_size_in_bits, biggest_align_in_bits, 0, union_inner_di_types,
gen_field_count, 0, "");
uint64_t union_offset_in_bits = 8*LLVMOffsetOfElement(g->target_data_ref, union_type->type_ref,
union_type->data.unionation.gen_union_index);
uint64_t tag_offset_in_bits = 8*LLVMOffsetOfElement(g->target_data_ref, union_type->type_ref,
union_type->data.unionation.gen_tag_index);
ZigLLVMDIType *union_member_di_type = ZigLLVMCreateDebugMemberType(g->dbuilder,
ZigLLVMTypeToScope(union_type->di_type), "union_field",
import->di_file, (unsigned)(decl_node->line + 1),
biggest_size_in_bits,
biggest_align_in_bits,
union_offset_in_bits,
0, union_di_type);
ZigLLVMDIType *tag_member_di_type = ZigLLVMCreateDebugMemberType(g->dbuilder,
ZigLLVMTypeToScope(union_type->di_type), "tag_field",
import->di_file, (unsigned)(decl_node->line + 1),
tag_debug_size_in_bits,
tag_debug_align_in_bits,
tag_offset_in_bits,
0, tag_di_type);
ZigLLVMDIType *di_root_members[2];
di_root_members[union_type->data.unionation.gen_tag_index] = tag_member_di_type;
di_root_members[union_type->data.unionation.gen_union_index] = union_member_di_type;
uint64_t debug_size_in_bits = 8*LLVMStoreSizeOfType(g->target_data_ref, union_type->type_ref);
uint64_t debug_align_in_bits = 8*LLVMABISizeOfType(g->target_data_ref, union_type->type_ref);
ZigLLVMDIType *replacement_di_type = ZigLLVMCreateDebugStructType(g->dbuilder,
ZigLLVMFileToScope(import->di_file),
buf_ptr(&union_type->name),
import->di_file, (unsigned)(decl_node->line + 1),
debug_size_in_bits,
debug_align_in_bits,
0, nullptr, di_root_members, 2, 0, nullptr, "");
ZigLLVMReplaceTemporary(g->dbuilder, union_type->di_type, replacement_di_type);
union_type->di_type = replacement_di_type;
}
static void resolve_enum_zero_bits(CodeGen *g, TypeTableEntry *enum_type) {
@ -1873,7 +2113,7 @@ static void resolve_enum_zero_bits(CodeGen *g, TypeTableEntry *enum_type) {
type_enum_field->value = i;
type_ensure_zero_bits_known(g, field_type);
if (field_type->id == TypeTableEntryIdInvalid) {
if (type_is_invalid(field_type)) {
enum_type->data.enumeration.is_invalid = true;
continue;
}
@ -1980,7 +2220,69 @@ static void resolve_struct_zero_bits(CodeGen *g, TypeTableEntry *struct_type) {
}
static void resolve_union_zero_bits(CodeGen *g, TypeTableEntry *union_type) {
zig_panic("TODO resolve_union_zero_bits");
assert(union_type->id == TypeTableEntryIdUnion);
if (union_type->data.unionation.zero_bits_known)
return;
if (union_type->data.unionation.zero_bits_loop_flag) {
union_type->data.unionation.zero_bits_known = true;
return;
}
union_type->data.unionation.zero_bits_loop_flag = true;
AstNode *decl_node = union_type->data.unionation.decl_node;
assert(decl_node->type == NodeTypeContainerDecl);
assert(union_type->di_type);
assert(!union_type->data.unionation.fields);
uint32_t field_count = (uint32_t)decl_node->data.container_decl.fields.length;
union_type->data.unionation.src_field_count = field_count;
union_type->data.unionation.fields = allocate<TypeUnionField>(field_count);
uint32_t biggest_align_bytes = 0;
Scope *scope = &union_type->data.unionation.decls_scope->base;
uint32_t gen_field_index = 0;
for (uint32_t i = 0; i < field_count; i += 1) {
AstNode *field_node = decl_node->data.container_decl.fields.at(i);
TypeUnionField *type_union_field = &union_type->data.unionation.fields[i];
type_union_field->name = field_node->data.struct_field.name;
TypeTableEntry *field_type = analyze_type_expr(g, scope, field_node->data.struct_field.type);
type_union_field->type_entry = field_type;
type_union_field->value = i;
type_ensure_zero_bits_known(g, field_type);
if (type_is_invalid(field_type)) {
union_type->data.unionation.is_invalid = true;
continue;
}
if (!type_has_bits(field_type))
continue;
type_union_field->gen_index = gen_field_index;
gen_field_index += 1;
uint32_t field_align_bytes = get_abi_alignment(g, field_type);
if (field_align_bytes > biggest_align_bytes) {
biggest_align_bytes = field_align_bytes;
}
}
bool auto_layout = (union_type->data.unionation.layout == ContainerLayoutAuto);
union_type->data.unionation.zero_bits_loop_flag = false;
union_type->data.unionation.gen_field_count = gen_field_index;
union_type->zero_bits = (gen_field_index == 0 && (field_count < 2 || !auto_layout));
union_type->data.unionation.zero_bits_known = true;
// also compute abi_alignment
if (!union_type->zero_bits) {
union_type->data.unionation.abi_alignment = biggest_align_bytes;
}
}
static void get_fully_qualified_decl_name_internal(Buf *buf, Scope *scope, uint8_t sep) {
@ -2851,6 +3153,18 @@ TypeStructField *find_struct_type_field(TypeTableEntry *type_entry, Buf *name) {
return nullptr;
}
TypeUnionField *find_union_type_field(TypeTableEntry *type_entry, Buf *name) {
assert(type_entry->id == TypeTableEntryIdUnion);
assert(type_entry->data.unionation.complete);
for (uint32_t i = 0; i < type_entry->data.unionation.src_field_count; i += 1) {
TypeUnionField *field = &type_entry->data.unionation.fields[i];
if (buf_eql_buf(field->name, name)) {
return field;
}
}
return nullptr;
}
static bool is_container(TypeTableEntry *type_entry) {
switch (type_entry->id) {
case TypeTableEntryIdInvalid:
@ -4703,6 +5017,8 @@ ConstParent *get_const_val_parent(CodeGen *g, ConstExprValue *value) {
return &value->data.x_array.s_none.parent;
} else if (type_entry->id == TypeTableEntryIdStruct) {
return &value->data.x_struct.parent;
} else if (type_entry->id == TypeTableEntryIdUnion) {
return &value->data.x_union.parent;
}
return nullptr;
}
@ -4914,7 +5230,8 @@ uint32_t get_abi_alignment(CodeGen *g, TypeTableEntry *type_entry) {
assert(type_entry->data.enumeration.abi_alignment != 0);
return type_entry->data.enumeration.abi_alignment;
} else if (type_entry->id == TypeTableEntryIdUnion) {
zig_panic("TODO");
assert(type_entry->data.unionation.abi_alignment != 0);
return type_entry->data.unionation.abi_alignment;
} else if (type_entry->id == TypeTableEntryIdOpaque) {
return 1;
} else {
@ -4929,3 +5246,11 @@ TypeTableEntry *get_align_amt_type(CodeGen *g) {
}
return g->align_amt_type;
}
uint32_t type_ptr_hash(const TypeTableEntry *ptr) {
return hash_ptr((void*)ptr);
}
bool type_ptr_eql(const TypeTableEntry *a, const TypeTableEntry *b) {
return a == b;
}

1
src/analyze.hpp
View File

@ -63,6 +63,7 @@ void resolve_container_type(CodeGen *g, TypeTableEntry *type_entry);
TypeStructField *find_struct_type_field(TypeTableEntry *type_entry, Buf *name);
ScopeDecls *get_container_scope(TypeTableEntry *type_entry);
TypeEnumField *find_enum_type_field(TypeTableEntry *enum_type, Buf *name);
TypeUnionField *find_union_type_field(TypeTableEntry *type_entry, Buf *name);
bool is_container_ref(TypeTableEntry *type_entry);
void scan_decls(CodeGen *g, ScopeDecls *decls_scope, AstNode *node);
void scan_import(CodeGen *g, ImportTableEntry *import);

8
src/c_tokenizer.cpp
View File

@ -120,6 +120,7 @@ static void begin_token(CTokenize *ctok, CTokId id) {
case CTokIdLParen:
case CTokIdRParen:
case CTokIdEOF:
case CTokIdDot:
break;
}
}
@ -216,9 +217,8 @@ void tokenize_c_macro(CTokenize *ctok, const uint8_t *c) {
buf_append_char(&ctok->buf, '0');
break;
case '.':
begin_token(ctok, CTokIdNumLitFloat);
ctok->state = CTokStateFloat;
buf_init_from_str(&ctok->buf, "0.");
begin_token(ctok, CTokIdDot);
end_token(ctok);
break;
case '(':
begin_token(ctok, CTokIdLParen);
@ -238,6 +238,8 @@ void tokenize_c_macro(CTokenize *ctok, const uint8_t *c) {
break;
case CTokStateFloat:
switch (*c) {
case '.':
break;
case 'e':
case 'E':
buf_append_char(&ctok->buf, 'e');

1
src/c_tokenizer.hpp
View File

@ -21,6 +21,7 @@ enum CTokId {
CTokIdLParen,
CTokIdRParen,
CTokIdEOF,
CTokIdDot,
};
enum CNumLitSuffix {

161
src/codegen.cpp
View File

@ -15,7 +15,7 @@
#include "ir.hpp"
#include "link.hpp"
#include "os.hpp"
#include "parsec.hpp"
#include "translate_c.hpp"
#include "target.hpp"
#include "zig_llvm.hpp"
@ -810,6 +810,8 @@ static Buf *panic_msg_buf(PanicMsgId msg_id) {
return buf_create_from_str("invalid error code");
case PanicMsgIdIncorrectAlignment:
return buf_create_from_str("incorrect alignment");
case PanicMsgIdBadUnionField:
return buf_create_from_str("access of inactive union field");
}
zig_unreachable();
}
@ -2393,6 +2395,50 @@ static LLVMValueRef ir_render_enum_field_ptr(CodeGen *g, IrExecutable *executabl
return bitcasted_union_field_ptr;
}
static LLVMValueRef ir_render_union_field_ptr(CodeGen *g, IrExecutable *executable,
IrInstructionUnionFieldPtr *instruction)
{
TypeTableEntry *union_ptr_type = instruction->union_ptr->value.type;
assert(union_ptr_type->id == TypeTableEntryIdPointer);
TypeTableEntry *union_type = union_ptr_type->data.pointer.child_type;
assert(union_type->id == TypeTableEntryIdUnion);
TypeUnionField *field = instruction->field;
if (!type_has_bits(field->type_entry))
return nullptr;
LLVMValueRef union_ptr = ir_llvm_value(g, instruction->union_ptr);
LLVMTypeRef field_type_ref = LLVMPointerType(field->type_entry->type_ref, 0);
if (union_type->data.unionation.gen_tag_index == SIZE_MAX) {
LLVMValueRef union_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, 0, "");
LLVMValueRef bitcasted_union_field_ptr = LLVMBuildBitCast(g->builder, union_field_ptr, field_type_ref, "");
return bitcasted_union_field_ptr;
}
if (ir_want_debug_safety(g, &instruction->base)) {
LLVMValueRef tag_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_tag_index, "");
LLVMValueRef tag_value = gen_load_untyped(g, tag_field_ptr, 0, false, "");
LLVMValueRef expected_tag_value = LLVMConstInt(union_type->data.unionation.tag_type->type_ref,
field->value, false);
LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnionCheckOk");
LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnionCheckFail");
LLVMValueRef ok_val = LLVMBuildICmp(g->builder, LLVMIntEQ, tag_value, expected_tag_value, "");
LLVMBuildCondBr(g->builder, ok_val, ok_block, bad_block);
LLVMPositionBuilderAtEnd(g->builder, bad_block);
gen_debug_safety_crash(g, PanicMsgIdBadUnionField);
LLVMPositionBuilderAtEnd(g->builder, ok_block);
}
LLVMValueRef union_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_union_index, "");
LLVMValueRef bitcasted_union_field_ptr = LLVMBuildBitCast(g->builder, union_field_ptr, field_type_ref, "");
return bitcasted_union_field_ptr;
}
static size_t find_asm_index(CodeGen *g, AstNode *node, AsmToken *tok) {
const char *ptr = buf_ptr(node->data.asm_expr.asm_template) + tok->start + 2;
size_t len = tok->end - tok->start - 2;
@ -3365,6 +3411,42 @@ static LLVMValueRef ir_render_struct_init(CodeGen *g, IrExecutable *executable,
return instruction->tmp_ptr;
}
static LLVMValueRef ir_render_union_init(CodeGen *g, IrExecutable *executable, IrInstructionUnionInit *instruction) {
TypeUnionField *type_union_field = instruction->field;
if (!type_has_bits(type_union_field->type_entry))
return nullptr;
uint32_t field_align_bytes = get_abi_alignment(g, type_union_field->type_entry);
TypeTableEntry *ptr_type = get_pointer_to_type_extra(g, type_union_field->type_entry,
false, false, field_align_bytes,
0, 0);
LLVMValueRef uncasted_union_ptr;
// Even if safety is off in this block, if the union type has the safety field, we have to populate it
// correctly. Otherwise safety code somewhere other than here could fail.
TypeTableEntry *union_type = instruction->union_type;
if (union_type->data.unionation.gen_tag_index != SIZE_MAX) {
LLVMValueRef tag_field_ptr = LLVMBuildStructGEP(g->builder, instruction->tmp_ptr,
union_type->data.unionation.gen_tag_index, "");
LLVMValueRef tag_value = LLVMConstInt(union_type->data.unionation.tag_type->type_ref,
type_union_field->value, false);
gen_store_untyped(g, tag_value, tag_field_ptr, 0, false);
uncasted_union_ptr = LLVMBuildStructGEP(g->builder, instruction->tmp_ptr,
(unsigned)union_type->data.unionation.gen_union_index, "");
} else {
uncasted_union_ptr = LLVMBuildStructGEP(g->builder, instruction->tmp_ptr, (unsigned)0, "");
}
LLVMValueRef field_ptr = LLVMBuildBitCast(g->builder, uncasted_union_ptr, ptr_type->type_ref, "");
LLVMValueRef value = ir_llvm_value(g, instruction->init_value);
gen_assign_raw(g, field_ptr, ptr_type, value);
return instruction->tmp_ptr;
}
static LLVMValueRef ir_render_container_init_list(CodeGen *g, IrExecutable *executable,
IrInstructionContainerInitList *instruction)
{
@ -3486,6 +3568,8 @@ static LLVMValueRef ir_render_instruction(CodeGen *g, IrExecutable *executable,
return ir_render_struct_field_ptr(g, executable, (IrInstructionStructFieldPtr *)instruction);
case IrInstructionIdEnumFieldPtr:
return ir_render_enum_field_ptr(g, executable, (IrInstructionEnumFieldPtr *)instruction);
case IrInstructionIdUnionFieldPtr:
return ir_render_union_field_ptr(g, executable, (IrInstructionUnionFieldPtr *)instruction);
case IrInstructionIdAsm:
return ir_render_asm(g, executable, (IrInstructionAsm *)instruction);
case IrInstructionIdTestNonNull:
@ -3544,6 +3628,8 @@ static LLVMValueRef ir_render_instruction(CodeGen *g, IrExecutable *executable,
return ir_render_init_enum(g, executable, (IrInstructionInitEnum *)instruction);
case IrInstructionIdStructInit:
return ir_render_struct_init(g, executable, (IrInstructionStructInit *)instruction);
case IrInstructionIdUnionInit:
return ir_render_union_init(g, executable, (IrInstructionUnionInit *)instruction);
case IrInstructionIdPtrCast:
return ir_render_ptr_cast(g, executable, (IrInstructionPtrCast *)instruction);
case IrInstructionIdBitCast:
@ -3595,6 +3681,7 @@ static void ir_render(CodeGen *g, FnTableEntry *fn_entry) {
static LLVMValueRef gen_const_ptr_struct_recursive(CodeGen *g, ConstExprValue *struct_const_val, size_t field_index);
static LLVMValueRef gen_const_ptr_array_recursive(CodeGen *g, ConstExprValue *array_const_val, size_t index);
static LLVMValueRef gen_const_ptr_union_recursive(CodeGen *g, ConstExprValue *array_const_val);
static LLVMValueRef gen_parent_ptr(CodeGen *g, ConstExprValue *val, ConstParent *parent) {
switch (parent->id) {
@ -3608,6 +3695,8 @@ static LLVMValueRef gen_parent_ptr(CodeGen *g, ConstExprValue *val, ConstParent
case ConstParentIdArray:
return gen_const_ptr_array_recursive(g, parent->data.p_array.array_val,
parent->data.p_array.elem_index);
case ConstParentIdUnion:
return gen_const_ptr_union_recursive(g, parent->data.p_union.union_val);
}
zig_unreachable();
}
@ -3637,6 +3726,18 @@ static LLVMValueRef gen_const_ptr_struct_recursive(CodeGen *g, ConstExprValue *s
return LLVMConstInBoundsGEP(base_ptr, indices, 2);
}
static LLVMValueRef gen_const_ptr_union_recursive(CodeGen *g, ConstExprValue *union_const_val) {
ConstParent *parent = &union_const_val->data.x_union.parent;
LLVMValueRef base_ptr = gen_parent_ptr(g, union_const_val, parent);
TypeTableEntry *u32 = g->builtin_types.entry_u32;
LLVMValueRef indices[] = {
LLVMConstNull(u32->type_ref),
LLVMConstInt(u32->type_ref, 0, false),
};
return LLVMConstInBoundsGEP(base_ptr, indices, 2);
}
static LLVMValueRef pack_const_int(CodeGen *g, LLVMTypeRef big_int_type_ref, ConstExprValue *const_val) {
switch (const_val->special) {
case ConstValSpecialRuntime:
@ -3872,10 +3973,6 @@ static LLVMValueRef gen_const_val(CodeGen *g, ConstExprValue *const_val) {
return LLVMConstNamedStruct(type_entry->type_ref, fields, type_entry->data.structure.gen_field_count);
}
}
case TypeTableEntryIdUnion:
{
zig_panic("TODO");
}
case TypeTableEntryIdArray:
{
uint64_t len = type_entry->data.array.len;
@ -3898,6 +3995,55 @@ static LLVMValueRef gen_const_val(CodeGen *g, ConstExprValue *const_val) {
return LLVMConstArray(element_type_ref, values, (unsigned)len);
}
}
case TypeTableEntryIdUnion:
{
LLVMTypeRef union_type_ref = type_entry->data.unionation.union_type_ref;
ConstExprValue *payload_value = const_val->data.x_union.payload;
assert(payload_value != nullptr);
if (!type_has_bits(payload_value->type)) {
return LLVMGetUndef(union_type_ref);
}
uint64_t field_type_bytes = LLVMStoreSizeOfType(g->target_data_ref, payload_value->type->type_ref);
uint64_t pad_bytes = type_entry->data.unionation.union_size_bytes - field_type_bytes;
LLVMValueRef correctly_typed_value = gen_const_val(g, payload_value);
bool make_unnamed_struct = is_llvm_value_unnamed_type(payload_value->type, correctly_typed_value) ||
payload_value->type != type_entry->data.unionation.most_aligned_union_member;
LLVMValueRef union_value_ref;
{
if (pad_bytes == 0) {
union_value_ref = correctly_typed_value;
} else {
LLVMValueRef fields[2];
fields[0] = correctly_typed_value;
fields[1] = LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), (unsigned)pad_bytes));
if (make_unnamed_struct || type_entry->data.unionation.gen_tag_index != SIZE_MAX) {
union_value_ref = LLVMConstStruct(fields, 2, false);
} else {
union_value_ref = LLVMConstNamedStruct(union_type_ref, fields, 2);
}
}
}
if (type_entry->data.unionation.gen_tag_index == SIZE_MAX) {
return union_value_ref;
}
LLVMValueRef tag_value = LLVMConstInt(type_entry->data.unionation.tag_type->type_ref, const_val->data.x_union.tag, false);
LLVMValueRef fields[2];
fields[type_entry->data.unionation.gen_union_index] = union_value_ref;
fields[type_entry->data.unionation.gen_tag_index] = tag_value;
if (make_unnamed_struct) {
return LLVMConstStruct(fields, 2, false);
} else {
return LLVMConstNamedStruct(type_entry->type_ref, fields, 2);
}
}
case TypeTableEntryIdEnum:
{
LLVMTypeRef tag_type_ref = type_entry->data.enumeration.tag_type->type_ref;
@ -4376,6 +4522,9 @@ static void do_code_gen(CodeGen *g) {
} else if (instruction->id == IrInstructionIdStructInit) {
IrInstructionStructInit *struct_init_instruction = (IrInstructionStructInit *)instruction;
slot = &struct_init_instruction->tmp_ptr;
} else if (instruction->id == IrInstructionIdUnionInit) {
IrInstructionUnionInit *union_init_instruction = (IrInstructionUnionInit *)instruction;
slot = &union_init_instruction->tmp_ptr;
} else if (instruction->id == IrInstructionIdCall) {
IrInstructionCall *call_instruction = (IrInstructionCall *)instruction;
slot = &call_instruction->tmp_ptr;
@ -5204,7 +5353,7 @@ static void init(CodeGen *g) {
define_builtin_compile_vars(g);
}
void codegen_parsec(CodeGen *g, Buf *full_path) {
void codegen_translate_c(CodeGen *g, Buf *full_path) {
find_libc_include_path(g);
Buf *src_basename = buf_alloc();

2
src/codegen.hpp
View File

@ -56,7 +56,7 @@ PackageTableEntry *codegen_create_package(CodeGen *g, const char *root_src_dir,
void codegen_add_assembly(CodeGen *g, Buf *path);
void codegen_add_object(CodeGen *g, Buf *object_path);
void codegen_parsec(CodeGen *g, Buf *path);
void codegen_translate_c(CodeGen *g, Buf *path);
#endif

261
src/ir.cpp
View File

@ -11,7 +11,7 @@
#include "ir.hpp"
#include "ir_print.hpp"
#include "os.hpp"
#include "parsec.hpp"
#include "translate_c.hpp"
#include "range_set.hpp"
#include "softfloat.hpp"
@ -227,6 +227,10 @@ static constexpr IrInstructionId ir_instruction_id(IrInstructionEnumFieldPtr *)
return IrInstructionIdEnumFieldPtr;
}
static constexpr IrInstructionId ir_instruction_id(IrInstructionUnionFieldPtr *) {
return IrInstructionIdUnionFieldPtr;
}
static constexpr IrInstructionId ir_instruction_id(IrInstructionElemPtr *) {
return IrInstructionIdElemPtr;
}
@ -351,6 +355,10 @@ static constexpr IrInstructionId ir_instruction_id(IrInstructionStructInit *) {
return IrInstructionIdStructInit;
}
static constexpr IrInstructionId ir_instruction_id(IrInstructionUnionInit *) {
return IrInstructionIdUnionInit;
}
static constexpr IrInstructionId ir_instruction_id(IrInstructionMinValue *) {
return IrInstructionIdMinValue;
}
@ -922,6 +930,27 @@ static IrInstruction *ir_build_enum_field_ptr_from(IrBuilder *irb, IrInstruction
return new_instruction;
}
static IrInstruction *ir_build_union_field_ptr(IrBuilder *irb, Scope *scope, AstNode *source_node,
IrInstruction *union_ptr, TypeUnionField *field)
{
IrInstructionUnionFieldPtr *instruction = ir_build_instruction<IrInstructionUnionFieldPtr>(irb, scope, source_node);
instruction->union_ptr = union_ptr;
instruction->field = field;
ir_ref_instruction(union_ptr, irb->current_basic_block);
return &instruction->base;
}
static IrInstruction *ir_build_union_field_ptr_from(IrBuilder *irb, IrInstruction *old_instruction,
IrInstruction *union_ptr, TypeUnionField *type_union_field)
{
IrInstruction *new_instruction = ir_build_union_field_ptr(irb, old_instruction->scope,
old_instruction->source_node, union_ptr, type_union_field);
ir_link_new_instruction(new_instruction, old_instruction);
return new_instruction;
}
static IrInstruction *ir_build_call(IrBuilder *irb, Scope *scope, AstNode *source_node,
FnTableEntry *fn_entry, IrInstruction *fn_ref, size_t arg_count, IrInstruction **args,
bool is_comptime, bool is_inline)
@ -1112,6 +1141,28 @@ static IrInstruction *ir_build_struct_init_from(IrBuilder *irb, IrInstruction *o
return new_instruction;
}
static IrInstruction *ir_build_union_init(IrBuilder *irb, Scope *scope, AstNode *source_node,
TypeTableEntry *union_type, TypeUnionField *field, IrInstruction *init_value)
{
IrInstructionUnionInit *union_init_instruction = ir_build_instruction<IrInstructionUnionInit>(irb, scope, source_node);
union_init_instruction->union_type = union_type;
union_init_instruction->field = field;
union_init_instruction->init_value = init_value;
ir_ref_instruction(init_value, irb->current_basic_block);
return &union_init_instruction->base;
}
static IrInstruction *ir_build_union_init_from(IrBuilder *irb, IrInstruction *old_instruction,
TypeTableEntry *union_type, TypeUnionField *field, IrInstruction *init_value)
{
IrInstruction *new_instruction = ir_build_union_init(irb, old_instruction->scope,
old_instruction->source_node, union_type, field, init_value);
ir_link_new_instruction(new_instruction, old_instruction);
return new_instruction;
}
static IrInstruction *ir_build_unreachable(IrBuilder *irb, Scope *scope, AstNode *source_node) {
IrInstructionUnreachable *unreachable_instruction =
ir_build_instruction<IrInstructionUnreachable>(irb, scope, source_node);
@ -2422,6 +2473,13 @@ static IrInstruction *ir_instruction_enumfieldptr_get_dep(IrInstructionEnumField
}
}
static IrInstruction *ir_instruction_unionfieldptr_get_dep(IrInstructionUnionFieldPtr *instruction, size_t index) {
switch (index) {
case 0: return instruction->union_ptr;
default: return nullptr;
}
}
static IrInstruction *ir_instruction_elemptr_get_dep(IrInstructionElemPtr *instruction, size_t index) {
switch (index) {
case 0: return instruction->array_ptr;
@ -2485,6 +2543,13 @@ static IrInstruction *ir_instruction_structinit_get_dep(IrInstructionStructInit
return nullptr;
}
static IrInstruction *ir_instruction_unioninit_get_dep(IrInstructionUnionInit *instruction, size_t index) {
switch (index) {
case 0: return instruction->init_value;
default: return nullptr;
}
}
static IrInstruction *ir_instruction_unreachable_get_dep(IrInstructionUnreachable *instruction, size_t index) {
return nullptr;
}
@ -3099,6 +3164,8 @@ static IrInstruction *ir_instruction_get_dep(IrInstruction *instruction, size_t
return ir_instruction_structfieldptr_get_dep((IrInstructionStructFieldPtr *) instruction, index);
case IrInstructionIdEnumFieldPtr:
return ir_instruction_enumfieldptr_get_dep((IrInstructionEnumFieldPtr *) instruction, index);
case IrInstructionIdUnionFieldPtr:
return ir_instruction_unionfieldptr_get_dep((IrInstructionUnionFieldPtr *) instruction, index);
case IrInstructionIdElemPtr:
return ir_instruction_elemptr_get_dep((IrInstructionElemPtr *) instruction, index);
case IrInstructionIdVarPtr:
@ -3117,6 +3184,8 @@ static IrInstruction *ir_instruction_get_dep(IrInstruction *instruction, size_t
return ir_instruction_containerinitfields_get_dep((IrInstructionContainerInitFields *) instruction, index);
case IrInstructionIdStructInit:
return ir_instruction_structinit_get_dep((IrInstructionStructInit *) instruction, index);
case IrInstructionIdUnionInit:
return ir_instruction_unioninit_get_dep((IrInstructionUnionInit *) instruction, index);
case IrInstructionIdUnreachable:
return ir_instruction_unreachable_get_dep((IrInstructionUnreachable *) instruction, index);
case IrInstructionIdTypeOf:
@ -6233,8 +6302,9 @@ static Buf *get_anon_type_name(CodeGen *codegen, IrExecutable *exec, const char
buf_appendf(name, ")");
return name;
} else {
//Note: C-imports do not have valid location information
return buf_sprintf("(anonymous %s at %s:%" ZIG_PRI_usize ":%" ZIG_PRI_usize ")", kind_name,
buf_ptr(source_node->owner->path), source_node->line + 1, source_node->column + 1);
(source_node->owner->path != nullptr) ? buf_ptr(source_node->owner->path) : "(null)", source_node->line + 1, source_node->column + 1);
}
}
}
@ -6263,6 +6333,9 @@ static IrInstruction *ir_gen_container_decl(IrBuilder *irb, Scope *parent_scope,
}
irb->codegen->resolve_queue.append(&tld_container->base);
// Add this to the list to mark as invalid if analyzing this exec fails.
irb->exec->tld_list.append(&tld_container->base);
return ir_build_const_type(irb, parent_scope, node, container_type);
}
@ -6485,6 +6558,20 @@ static bool ir_goto_pass2(IrBuilder *irb) {
return true;
}
static void invalidate_exec(IrExecutable *exec) {
if (exec->invalid)
return;
exec->invalid = true;
for (size_t i = 0; i < exec->tld_list.length; i += 1) {
exec->tld_list.items[i]->resolution = TldResolutionInvalid;
}
if (exec->source_exec != nullptr)
invalidate_exec(exec->source_exec);
}
bool ir_gen(CodeGen *codegen, AstNode *node, Scope *scope, IrExecutable *ir_executable) {
assert(node->owner);
@ -6508,7 +6595,7 @@ bool ir_gen(CodeGen *codegen, AstNode *node, Scope *scope, IrExecutable *ir_exec
}
if (!ir_goto_pass2(irb)) {
irb->exec->invalid = true;
invalidate_exec(ir_executable);
return false;
}
@ -6534,7 +6621,7 @@ static void add_call_stack_errors(CodeGen *codegen, IrExecutable *exec, ErrorMsg
}
static ErrorMsg *exec_add_error_node(CodeGen *codegen, IrExecutable *exec, AstNode *source_node, Buf *msg) {
exec->invalid = true;
invalidate_exec(exec);
ErrorMsg *err_msg = add_node_error(codegen, source_node, msg);
if (exec->parent_exec) {
add_call_stack_errors(codegen, exec, err_msg, 10);
@ -7897,6 +7984,9 @@ static IrInstruction *ir_get_const_ptr(IrAnalyze *ira, IrInstruction *instructio
ConstExprValue *const_val = &const_instr->value;
const_val->type = pointee_type;
type_ensure_zero_bits_known(ira->codegen, type_entry);
if (type_is_invalid(type_entry)) {
return ira->codegen->invalid_instruction;
}
const_val->data.x_type = get_pointer_to_type_extra(ira->codegen, type_entry,
ptr_is_const, ptr_is_volatile, get_abi_alignment(ira->codegen, type_entry), 0, 0);
return const_instr;
@ -7984,6 +8074,7 @@ IrInstruction *ir_eval_const_value(CodeGen *codegen, Scope *scope, AstNode *node
IrExecutable analyzed_executable = {0};
analyzed_executable.source_node = source_node;
analyzed_executable.parent_exec = parent_exec;
analyzed_executable.source_exec = &ir_executable;
analyzed_executable.name = exec_name;
analyzed_executable.is_inline = true;
analyzed_executable.fn_entry = fn_entry;
@ -10279,30 +10370,40 @@ static IrInstruction *ir_get_var_ptr(IrAnalyze *ira, IrInstruction *instruction,
bool is_const = (var->value->type->id == TypeTableEntryIdMetaType) ? is_const_ptr : var->src_is_const;
bool is_volatile = (var->value->type->id == TypeTableEntryIdMetaType) ? is_volatile_ptr : false;
if (mem_slot && mem_slot->special != ConstValSpecialRuntime) {
ConstPtrMut ptr_mut;
if (comptime_var_mem) {
ptr_mut = ConstPtrMutComptimeVar;
} else if (var->gen_is_const) {
ptr_mut = ConstPtrMutComptimeConst;
} else {
assert(!comptime_var_mem);
ptr_mut = ConstPtrMutRuntimeVar;
if (mem_slot != nullptr) {
switch (mem_slot->special) {
case ConstValSpecialRuntime:
goto no_mem_slot;
case ConstValSpecialStatic: // fallthrough
case ConstValSpecialUndef: {
ConstPtrMut ptr_mut;
if (comptime_var_mem) {
ptr_mut = ConstPtrMutComptimeVar;
} else if (var->gen_is_const) {
ptr_mut = ConstPtrMutComptimeConst;
} else {
assert(!comptime_var_mem);
ptr_mut = ConstPtrMutRuntimeVar;
}
return ir_get_const_ptr(ira, instruction, mem_slot, var->value->type,
ptr_mut, is_const, is_volatile, var->align_bytes);
}
}
return ir_get_const_ptr(ira, instruction, mem_slot, var->value->type,
ptr_mut, is_const, is_volatile, var->align_bytes);
} else {
IrInstruction *var_ptr_instruction = ir_build_var_ptr(&ira->new_irb,
instruction->scope, instruction->source_node, var, is_const, is_volatile);
var_ptr_instruction->value.type = get_pointer_to_type_extra(ira->codegen, var->value->type,
var->src_is_const, is_volatile, var->align_bytes, 0, 0);
type_ensure_zero_bits_known(ira->codegen, var->value->type);
bool in_fn_scope = (scope_fn_entry(var->parent_scope) != nullptr);
var_ptr_instruction->value.data.rh_ptr = in_fn_scope ? RuntimeHintPtrStack : RuntimeHintPtrNonStack;
return var_ptr_instruction;
zig_unreachable();
}
no_mem_slot:
IrInstruction *var_ptr_instruction = ir_build_var_ptr(&ira->new_irb,
instruction->scope, instruction->source_node, var, is_const, is_volatile);
var_ptr_instruction->value.type = get_pointer_to_type_extra(ira->codegen, var->value->type,
var->src_is_const, is_volatile, var->align_bytes, 0, 0);
type_ensure_zero_bits_known(ira->codegen, var->value->type);
bool in_fn_scope = (scope_fn_entry(var->parent_scope) != nullptr);
var_ptr_instruction->value.data.rh_ptr = in_fn_scope ? RuntimeHintPtrStack : RuntimeHintPtrNonStack;
return var_ptr_instruction;
}
static TypeTableEntry *ir_analyze_fn_call(IrAnalyze *ira, IrInstructionCall *call_instruction,
@ -10408,10 +10509,11 @@ static TypeTableEntry *ir_analyze_fn_call(IrAnalyze *ira, IrInstructionCall *cal
result = ir_eval_const_value(ira->codegen, exec_scope, body_node, return_type,
ira->new_irb.exec->backward_branch_count, ira->new_irb.exec->backward_branch_quota, fn_entry,
nullptr, call_instruction->base.source_node, nullptr, ira->new_irb.exec);
if (type_is_invalid(result->value.type))
return ira->codegen->builtin_types.entry_invalid;
ira->codegen->memoized_fn_eval_table.put(exec_scope, result);
if (type_is_invalid(result->value.type))
return ira->codegen->builtin_types.entry_invalid;
}
ConstExprValue *out_val = ir_build_const_from(ira, &call_instruction->base);
@ -11417,8 +11519,20 @@ static TypeTableEntry *ir_analyze_container_member_access_inner(IrAnalyze *ira,
return ir_analyze_ref(ira, &field_ptr_instruction->base, bound_fn_value, true, false);
}
}
const char *prefix_name;
if (is_slice(bare_struct_type)) {
prefix_name = "";
} else if (bare_struct_type->id == TypeTableEntryIdStruct) {
prefix_name = "struct ";
} else if (bare_struct_type->id == TypeTableEntryIdEnum) {
prefix_name = "enum ";
} else if (bare_struct_type->id == TypeTableEntryIdUnion) {
prefix_name = "union ";
} else {
prefix_name = "";
}
ir_add_error_node(ira, field_ptr_instruction->base.source_node,
buf_sprintf("no member named '%s' in '%s'", buf_ptr(field_name), buf_ptr(&bare_struct_type->name)));
buf_sprintf("no member named '%s' in %s'%s'", buf_ptr(field_name), prefix_name, buf_ptr(&bare_struct_type->name)));
return ira->codegen->builtin_types.entry_invalid;
}
@ -11428,14 +11542,13 @@ static TypeTableEntry *ir_analyze_container_field_ptr(IrAnalyze *ira, Buf *field
{
TypeTableEntry *bare_type = container_ref_type(container_type);
ensure_complete_type(ira->codegen, bare_type);
if (type_is_invalid(bare_type))
return ira->codegen->builtin_types.entry_invalid;
assert(container_ptr->value.type->id == TypeTableEntryIdPointer);
bool is_const = container_ptr->value.type->data.pointer.is_const;
bool is_volatile = container_ptr->value.type->data.pointer.is_volatile;
if (bare_type->id == TypeTableEntryIdStruct) {
if (bare_type->data.structure.is_invalid)
return ira->codegen->builtin_types.entry_invalid;
TypeStructField *field = find_struct_type_field(bare_type, field_name);
if (field) {
bool is_packed = (bare_type->data.structure.layout == ContainerLayoutPacked);
@ -11476,9 +11589,6 @@ static TypeTableEntry *ir_analyze_container_field_ptr(IrAnalyze *ira, Buf *field
field_ptr_instruction, container_ptr, container_type);
}
} else if (bare_type->id == TypeTableEntryIdEnum) {
if (bare_type->data.enumeration.is_invalid)
return ira->codegen->builtin_types.entry_invalid;
TypeEnumField *field = find_enum_type_field(bare_type, field_name);
if (field) {
ir_build_enum_field_ptr_from(&ira->new_irb, &field_ptr_instruction->base, container_ptr, field);
@ -11489,7 +11599,15 @@ static TypeTableEntry *ir_analyze_container_field_ptr(IrAnalyze *ira, Buf *field
field_ptr_instruction, container_ptr, container_type);
}
} else if (bare_type->id == TypeTableEntryIdUnion) {
zig_panic("TODO");
TypeUnionField *field = find_union_type_field(bare_type, field_name);
if (field) {
ir_build_union_field_ptr_from(&ira->new_irb, &field_ptr_instruction->base, container_ptr, field);
return get_pointer_to_type_extra(ira->codegen, field->type_entry, is_const, is_volatile,
get_abi_alignment(ira->codegen, field->type_entry), 0, 0);
} else {
return ir_analyze_container_member_access_inner(ira, bare_type, field_name,
field_ptr_instruction, container_ptr, container_type);
}
} else {
zig_unreachable();
}
@ -13033,9 +13151,71 @@ static TypeTableEntry *ir_analyze_instruction_ref(IrAnalyze *ira, IrInstructionR
return ir_analyze_ref(ira, &ref_instruction->base, value, ref_instruction->is_const, ref_instruction->is_volatile);
}
static TypeTableEntry *ir_analyze_container_init_fields_union(IrAnalyze *ira, IrInstruction *instruction,
TypeTableEntry *container_type, size_t instr_field_count, IrInstructionContainerInitFieldsField *fields)
{
assert(container_type->id == TypeTableEntryIdUnion);
ensure_complete_type(ira->codegen, container_type);
if (instr_field_count != 1) {
ir_add_error(ira, instruction,
buf_sprintf("union initialization expects exactly one field"));
return ira->codegen->builtin_types.entry_invalid;
}
IrInstructionContainerInitFieldsField *field = &fields[0];
IrInstruction *field_value = field->value->other;
if (type_is_invalid(field_value->value.type))
return ira->codegen->builtin_types.entry_invalid;
TypeUnionField *type_field = find_union_type_field(container_type, field->name);
if (!type_field) {
ir_add_error_node(ira, field->source_node,
buf_sprintf("no member named '%s' in union '%s'",
buf_ptr(field->name), buf_ptr(&container_type->name)));
return ira->codegen->builtin_types.entry_invalid;
}
if (type_is_invalid(type_field->type_entry))
return ira->codegen->builtin_types.entry_invalid;
IrInstruction *casted_field_value = ir_implicit_cast(ira, field_value, type_field->type_entry);
if (casted_field_value == ira->codegen->invalid_instruction)
return ira->codegen->builtin_types.entry_invalid;
bool is_comptime = ir_should_inline(ira->new_irb.exec, instruction->scope);
if (is_comptime || casted_field_value->value.special != ConstValSpecialRuntime) {
ConstExprValue *field_val = ir_resolve_const(ira, casted_field_value, UndefOk);
if (!field_val)
return ira->codegen->builtin_types.entry_invalid;
ConstExprValue *out_val = ir_build_const_from(ira, instruction);
out_val->data.x_union.payload = field_val;
out_val->data.x_union.tag = type_field->value;
ConstParent *parent = get_const_val_parent(ira->codegen, field_val);
if (parent != nullptr) {
parent->id = ConstParentIdUnion;
parent->data.p_union.union_val = out_val;
}
return container_type;
}
IrInstruction *new_instruction = ir_build_union_init_from(&ira->new_irb, instruction,
container_type, type_field, casted_field_value);
ir_add_alloca(ira, new_instruction, container_type);
return container_type;
}
static TypeTableEntry *ir_analyze_container_init_fields(IrAnalyze *ira, IrInstruction *instruction,
TypeTableEntry *container_type, size_t instr_field_count, IrInstructionContainerInitFieldsField *fields)
{
if (container_type->id == TypeTableEntryIdUnion) {
return ir_analyze_container_init_fields_union(ira, instruction, container_type, instr_field_count, fields);
}
if (container_type->id != TypeTableEntryIdStruct || is_slice(container_type)) {
ir_add_error(ira, instruction,
buf_sprintf("type '%s' does not support struct initialization syntax",
@ -13043,8 +13223,7 @@ static TypeTableEntry *ir_analyze_container_init_fields(IrAnalyze *ira, IrInstru
return ira->codegen->builtin_types.entry_invalid;
}
if (!type_is_complete(container_type))
resolve_container_type(ira->codegen, container_type);
ensure_complete_type(ira->codegen, container_type);
size_t actual_field_count = container_type->data.structure.src_field_count;
@ -13070,7 +13249,7 @@ static TypeTableEntry *ir_analyze_container_init_fields(IrAnalyze *ira, IrInstru
TypeStructField *type_field = find_struct_type_field(container_type, field->name);
if (!type_field) {
ir_add_error_node(ira, field->source_node,
buf_sprintf("no member named '%s' in '%s'",
buf_sprintf("no member named '%s' in struct '%s'",
buf_ptr(field->name), buf_ptr(&container_type->name)));
return ira->codegen->builtin_types.entry_invalid;
}
@ -15657,8 +15836,10 @@ static TypeTableEntry *ir_analyze_instruction_nocast(IrAnalyze *ira, IrInstructi
case IrInstructionIdIntToErr:
case IrInstructionIdErrToInt:
case IrInstructionIdStructInit:
case IrInstructionIdUnionInit:
case IrInstructionIdStructFieldPtr:
case IrInstructionIdEnumFieldPtr:
case IrInstructionIdUnionFieldPtr:
case IrInstructionIdInitEnum:
case IrInstructionIdMaybeWrap:
case IrInstructionIdErrWrapCode:
@ -15968,6 +16149,7 @@ bool ir_has_side_effects(IrInstruction *instruction) {
case IrInstructionIdContainerInitList:
case IrInstructionIdContainerInitFields:
case IrInstructionIdStructInit:
case IrInstructionIdUnionInit:
case IrInstructionIdFieldPtr:
case IrInstructionIdElemPtr:
case IrInstructionIdVarPtr:
@ -15977,6 +16159,7 @@ bool ir_has_side_effects(IrInstruction *instruction) {
case IrInstructionIdArrayLen:
case IrInstructionIdStructFieldPtr:
case IrInstructionIdEnumFieldPtr:
case IrInstructionIdUnionFieldPtr:
case IrInstructionIdArrayType:
case IrInstructionIdSliceType:
case IrInstructionIdSizeOf:

22
src/ir_print.cpp
View File

@ -290,6 +290,15 @@ static void ir_print_struct_init(IrPrint *irp, IrInstructionStructInit *instruct
fprintf(irp->f, "} // struct init");
}
static void ir_print_union_init(IrPrint *irp, IrInstructionUnionInit *instruction) {
Buf *field_name = instruction->field->name;
fprintf(irp->f, "%s {", buf_ptr(&instruction->union_type->name));
fprintf(irp->f, ".%s = ", buf_ptr(field_name));
ir_print_other_instruction(irp, instruction->init_value);
fprintf(irp->f, "} // union init");
}
static void ir_print_unreachable(IrPrint *irp, IrInstructionUnreachable *instruction) {
fprintf(irp->f, "unreachable");
}
@ -359,6 +368,13 @@ static void ir_print_enum_field_ptr(IrPrint *irp, IrInstructionEnumFieldPtr *ins
fprintf(irp->f, ")");
}
static void ir_print_union_field_ptr(IrPrint *irp, IrInstructionUnionFieldPtr *instruction) {
fprintf(irp->f, "@UnionFieldPtr(&");
ir_print_other_instruction(irp, instruction->union_ptr);
fprintf(irp->f, ".%s", buf_ptr(instruction->field->name));
fprintf(irp->f, ")");
}
static void ir_print_set_debug_safety(IrPrint *irp, IrInstructionSetDebugSafety *instruction) {
fprintf(irp->f, "@setDebugSafety(");
ir_print_other_instruction(irp, instruction->scope_value);
@ -1023,6 +1039,9 @@ static void ir_print_instruction(IrPrint *irp, IrInstruction *instruction) {
case IrInstructionIdStructInit:
ir_print_struct_init(irp, (IrInstructionStructInit *)instruction);
break;
case IrInstructionIdUnionInit:
ir_print_union_init(irp, (IrInstructionUnionInit *)instruction);
break;
case IrInstructionIdUnreachable:
ir_print_unreachable(irp, (IrInstructionUnreachable *)instruction);
break;
@ -1056,6 +1075,9 @@ static void ir_print_instruction(IrPrint *irp, IrInstruction *instruction) {
case IrInstructionIdEnumFieldPtr:
ir_print_enum_field_ptr(irp, (IrInstructionEnumFieldPtr *)instruction);
break;
case IrInstructionIdUnionFieldPtr:
ir_print_union_field_ptr(irp, (IrInstructionUnionFieldPtr *)instruction);
break;
case IrInstructionIdSetDebugSafety:
ir_print_set_debug_safety(irp, (IrInstructionSetDebugSafety *)instruction);
break;

22
src/main.cpp
View File

@ -23,7 +23,7 @@ static int usage(const char *arg0) {
" build-exe [source] create executable from source or object files\n"
" build-lib [source] create library from source or object files\n"
" build-obj [source] create object from source or assembly\n"
" parsec [source] convert c code to zig code\n"
" translate-c [source] convert c code to zig code\n"
" targets list available compilation targets\n"
" test [source] create and run a test build\n"
" version print version number and exit\n"
@ -229,7 +229,7 @@ enum Cmd {
CmdTest,
CmdVersion,
CmdZen,
CmdParseC,
CmdTranslateC,
CmdTargets,
};
@ -632,8 +632,8 @@ int main(int argc, char **argv) {
cmd = CmdVersion;
} else if (strcmp(arg, "zen") == 0) {
cmd = CmdZen;
} else if (strcmp(arg, "parsec") == 0) {
cmd = CmdParseC;
} else if (strcmp(arg, "translate-c") == 0) {
cmd = CmdTranslateC;
} else if (strcmp(arg, "test") == 0) {
cmd = CmdTest;
out_type = OutTypeExe;
@ -646,7 +646,7 @@ int main(int argc, char **argv) {
} else {
switch (cmd) {
case CmdBuild:
case CmdParseC:
case CmdTranslateC:
case CmdTest:
if (!in_file) {
in_file = arg;
@ -703,13 +703,13 @@ int main(int argc, char **argv) {
switch (cmd) {
case CmdBuild:
case CmdParseC:
case CmdTranslateC:
case CmdTest:
{
if (cmd == CmdBuild && !in_file && objects.length == 0 && asm_files.length == 0) {
fprintf(stderr, "Expected source file argument or at least one --object or --assembly argument.\n");
return usage(arg0);
} else if ((cmd == CmdParseC || cmd == CmdTest) && !in_file) {
} else if ((cmd == CmdTranslateC || cmd == CmdTest) && !in_file) {
fprintf(stderr, "Expected source file argument.\n");
return usage(arg0);
} else if (cmd == CmdBuild && out_type == OutTypeObj && objects.length != 0) {
@ -719,7 +719,7 @@ int main(int argc, char **argv) {
assert(cmd != CmdBuild || out_type != OutTypeUnknown);
bool need_name = (cmd == CmdBuild || cmd == CmdParseC);
bool need_name = (cmd == CmdBuild || cmd == CmdTranslateC);
Buf *in_file_buf = nullptr;
@ -742,7 +742,7 @@ int main(int argc, char **argv) {
return usage(arg0);
}
Buf *zig_root_source_file = (cmd == CmdParseC) ? nullptr : in_file_buf;
Buf *zig_root_source_file = (cmd == CmdTranslateC) ? nullptr : in_file_buf;
Buf *full_cache_dir = buf_alloc();
os_path_resolve(buf_create_from_str("."),
@ -841,8 +841,8 @@ int main(int argc, char **argv) {
if (timing_info)
codegen_print_timing_report(g, stdout);
return EXIT_SUCCESS;
} else if (cmd == CmdParseC) {
codegen_parsec(g, in_file_buf);
} else if (cmd == CmdTranslateC) {
codegen_translate_c(g, in_file_buf);
ast_render(g, stdout, g->root_import->root, 4);
if (timing_info)
codegen_print_timing_report(g, stdout);

2820
src/parsec.cpp → src/translate_c.cpp
View File

File diff suppressed because it is too large Load Diff

0
src/parsec.hpp → src/translate_c.hpp
View File

4
src/zig_llvm.cpp
View File

@ -403,6 +403,10 @@ unsigned ZigLLVMTag_DW_structure_type(void) {
return dwarf::DW_TAG_structure_type;
}
unsigned ZigLLVMTag_DW_union_type(void) {
return dwarf::DW_TAG_union_type;
}
ZigLLVMDIBuilder *ZigLLVMCreateDIBuilder(LLVMModuleRef module, bool allow_unresolved) {
DIBuilder *di_builder = new DIBuilder(*unwrap(module), allow_unresolved);
return reinterpret_cast<ZigLLVMDIBuilder *>(di_builder);

1
src/zig_llvm.hpp
View File

@ -117,6 +117,7 @@ unsigned ZigLLVMEncoding_DW_ATE_signed_char(void);
unsigned ZigLLVMLang_DW_LANG_C99(void);
unsigned ZigLLVMTag_DW_variable(void);
unsigned ZigLLVMTag_DW_structure_type(void);
unsigned ZigLLVMTag_DW_union_type(void);
ZigLLVMDIBuilder *ZigLLVMCreateDIBuilder(LLVMModuleRef module, bool allow_unresolved);
void ZigLLVMAddModuleDebugInfoFlag(LLVMModuleRef module);

583
std/base64.zig
View File

@ -1,186 +1,485 @@
const assert = @import("debug.zig").assert;
const mem = @import("mem.zig");
pub const standard_alphabet = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";
pub const standard_alphabet_chars = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";
pub const standard_pad_char = '=';
pub const standard_encoder = Base64Encoder.init(standard_alphabet_chars, standard_pad_char);
pub fn encode(dest: []u8, source: []const u8) -> []u8 {
return encodeWithAlphabet(dest, source, standard_alphabet);
}
pub const Base64Encoder = struct {
alphabet_chars: []const u8,
pad_char: u8,
/// invalid characters in source are allowed, but they cause the value of dest to be undefined.
pub fn decode(dest: []u8, source: []const u8) -> []u8 {
return decodeWithAlphabet(dest, source, standard_alphabet);
}
/// a bunch of assertions, then simply pass the data right through.
pub fn init(alphabet_chars: []const u8, pad_char: u8) -> Base64Encoder {
assert(alphabet_chars.len == 64);
var char_in_alphabet = []bool{false} ** 256;
for (alphabet_chars) |c| {
assert(!char_in_alphabet[c]);
assert(c != pad_char);
char_in_alphabet[c] = true;
}
pub fn encodeWithAlphabet(dest: []u8, source: []const u8, alphabet: []const u8) -> []u8 {
assert(alphabet.len == 65);
assert(dest.len >= calcEncodedSize(source.len));
var i: usize = 0;
var out_index: usize = 0;
while (i + 2 < source.len) : (i += 3) {
dest[out_index] = alphabet[(source[i] >> 2) & 0x3f];
out_index += 1;
dest[out_index] = alphabet[((source[i] & 0x3) << 4) |
((source[i + 1] & 0xf0) >> 4)];
out_index += 1;
dest[out_index] = alphabet[((source[i + 1] & 0xf) << 2) |
((source[i + 2] & 0xc0) >> 6)];
out_index += 1;
dest[out_index] = alphabet[source[i + 2] & 0x3f];
out_index += 1;
return Base64Encoder{
.alphabet_chars = alphabet_chars,
.pad_char = pad_char,
};
}
if (i < source.len) {
dest[out_index] = alphabet[(source[i] >> 2) & 0x3f];
out_index += 1;
/// ceil(source_len * 4/3)
pub fn calcSize(source_len: usize) -> usize {
return @divTrunc(source_len + 2, 3) * 4;
}
if (i + 1 == source.len) {
dest[out_index] = alphabet[(source[i] & 0x3) << 4];
/// dest.len must be what you get from ::calcSize.
pub fn encode(encoder: &const Base64Encoder, dest: []u8, source: []const u8) {
assert(dest.len == Base64Encoder.calcSize(source.len));
var i: usize = 0;
var out_index: usize = 0;
while (i + 2 < source.len) : (i += 3) {
dest[out_index] = encoder.alphabet_chars[(source[i] >> 2) & 0x3f];
out_index += 1;
dest[out_index] = alphabet[64];
out_index += 1;
} else {
dest[out_index] = alphabet[((source[i] & 0x3) << 4) |
dest[out_index] = encoder.alphabet_chars[((source[i] & 0x3) << 4) |
((source[i + 1] & 0xf0) >> 4)];
out_index += 1;
dest[out_index] = alphabet[(source[i + 1] & 0xf) << 2];
dest[out_index] = encoder.alphabet_chars[((source[i + 1] & 0xf) << 2) |
((source[i + 2] & 0xc0) >> 6)];
out_index += 1;
dest[out_index] = encoder.alphabet_chars[source[i + 2] & 0x3f];
out_index += 1;
}
dest[out_index] = alphabet[64];
out_index += 1;
if (i < source.len) {
dest[out_index] = encoder.alphabet_chars[(source[i] >> 2) & 0x3f];
out_index += 1;
if (i + 1 == source.len) {
dest[out_index] = encoder.alphabet_chars[(source[i] & 0x3) << 4];
out_index += 1;
dest[out_index] = encoder.pad_char;
out_index += 1;
} else {
dest[out_index] = encoder.alphabet_chars[((source[i] & 0x3) << 4) |
((source[i + 1] & 0xf0) >> 4)];
out_index += 1;
dest[out_index] = encoder.alphabet_chars[(source[i + 1] & 0xf) << 2];
out_index += 1;
}
dest[out_index] = encoder.pad_char;
out_index += 1;
}
}
};
pub const standard_decoder = Base64Decoder.init(standard_alphabet_chars, standard_pad_char);
error InvalidPadding;
error InvalidCharacter;
pub const Base64Decoder = struct {
/// e.g. 'A' => 0.
/// undefined for any value not in the 64 alphabet chars.
char_to_index: [256]u8,
/// true only for the 64 chars in the alphabet, not the pad char.
char_in_alphabet: [256]bool,
pad_char: u8,
pub fn init(alphabet_chars: []const u8, pad_char: u8) -> Base64Decoder {
assert(alphabet_chars.len == 64);
var result = Base64Decoder{
.char_to_index = undefined,
.char_in_alphabet = []bool{false} ** 256,
.pad_char = pad_char,
};
for (alphabet_chars) |c, i| {
assert(!result.char_in_alphabet[c]);
assert(c != pad_char);
result.char_to_index[c] = u8(i);
result.char_in_alphabet[c] = true;
}
return result;
}
return dest[0..out_index];
}
/// invalid characters in source are allowed, but they cause the value of dest to be undefined.
pub fn decodeWithAlphabet(dest: []u8, source: []const u8, alphabet: []const u8) -> []u8 {
assert(alphabet.len == 65);
var ascii6 = []u8{64} ** 256;
for (alphabet) |c, i| {
ascii6[c] = u8(i);
/// If the encoded buffer is detected to be invalid, returns error.InvalidPadding.
pub fn calcSize(decoder: &const Base64Decoder, source: []const u8) -> %usize {
if (source.len % 4 != 0) return error.InvalidPadding;
return calcDecodedSizeExactUnsafe(source, decoder.pad_char);
}
return decodeWithAscii6BitMap(dest, source, ascii6[0..], alphabet[64]);
}
/// dest.len must be what you get from ::calcSize.
/// invalid characters result in error.InvalidCharacter.
/// invalid padding results in error.InvalidPadding.
pub fn decode(decoder: &const Base64Decoder, dest: []u8, source: []const u8) -> %void {
assert(dest.len == %%decoder.calcSize(source));
assert(source.len % 4 == 0);
pub fn decodeWithAscii6BitMap(dest: []u8, source: []const u8, ascii6: []const u8, pad_char: u8) -> []u8 {
assert(ascii6.len == 256);
assert(dest.len >= calcExactDecodedSizeWithPadChar(source, pad_char));
var src_cursor: usize = 0;
var dest_cursor: usize = 0;
var src_index: usize = 0;
var dest_index: usize = 0;
var in_buf_len: usize = source.len;
while (src_cursor < source.len) : (src_cursor += 4) {
if (!decoder.char_in_alphabet[source[src_cursor + 0]]) return error.InvalidCharacter;
if (!decoder.char_in_alphabet[source[src_cursor + 1]]) return error.InvalidCharacter;
if (src_cursor < source.len - 4 or source[src_cursor + 3] != decoder.pad_char) {
// common case
if (!decoder.char_in_alphabet[source[src_cursor + 2]]) return error.InvalidCharacter;
if (!decoder.char_in_alphabet[source[src_cursor + 3]]) return error.InvalidCharacter;
dest[dest_cursor + 0] = decoder.char_to_index[source[src_cursor + 0]] << 2 |
decoder.char_to_index[source[src_cursor + 1]] >> 4;
dest[dest_cursor + 1] = decoder.char_to_index[source[src_cursor + 1]] << 4 |
decoder.char_to_index[source[src_cursor + 2]] >> 2;
dest[dest_cursor + 2] = decoder.char_to_index[source[src_cursor + 2]] << 6 |
decoder.char_to_index[source[src_cursor + 3]];
dest_cursor += 3;
} else if (source[src_cursor + 2] != decoder.pad_char) {
// one pad char
if (!decoder.char_in_alphabet[source[src_cursor + 2]]) return error.InvalidCharacter;
dest[dest_cursor + 0] = decoder.char_to_index[source[src_cursor + 0]] << 2 |
decoder.char_to_index[source[src_cursor + 1]] >> 4;
dest[dest_cursor + 1] = decoder.char_to_index[source[src_cursor + 1]] << 4 |
decoder.char_to_index[source[src_cursor + 2]] >> 2;
if (decoder.char_to_index[source[src_cursor + 2]] << 6 != 0) return error.InvalidPadding;
dest_cursor += 2;
} else {
// two pad chars
dest[dest_cursor + 0] = decoder.char_to_index[source[src_cursor + 0]] << 2 |
decoder.char_to_index[source[src_cursor + 1]] >> 4;
if (decoder.char_to_index[source[src_cursor + 1]] << 4 != 0) return error.InvalidPadding;
dest_cursor += 1;
}
}
while (in_buf_len > 0 and source[in_buf_len - 1] == pad_char) {
in_buf_len -= 1;
assert(src_cursor == source.len);
assert(dest_cursor == dest.len);
}
};
error OutputTooSmall;
pub const Base64DecoderWithIgnore = struct {
decoder: Base64Decoder,
char_is_ignored: [256]bool,
pub fn init(alphabet_chars: []const u8, pad_char: u8, ignore_chars: []const u8) -> Base64DecoderWithIgnore {
var result = Base64DecoderWithIgnore {
.decoder = Base64Decoder.init(alphabet_chars, pad_char),
.char_is_ignored = []bool{false} ** 256,
};
for (ignore_chars) |c| {
assert(!result.decoder.char_in_alphabet[c]);
assert(!result.char_is_ignored[c]);
assert(result.decoder.pad_char != c);
result.char_is_ignored[c] = true;
}
return result;
}
while (in_buf_len > 4) {
dest[dest_index] = ascii6[source[src_index + 0]] << 2 |
ascii6[source[src_index + 1]] >> 4;
dest_index += 1;
dest[dest_index] = ascii6[source[src_index + 1]] << 4 |
ascii6[source[src_index + 2]] >> 2;
dest_index += 1;
dest[dest_index] = ascii6[source[src_index + 2]] << 6 |
ascii6[source[src_index + 3]];
dest_index += 1;
src_index += 4;
in_buf_len -= 4;
/// If no characters end up being ignored or padding, this will be the exact decoded size.
pub fn calcSizeUpperBound(encoded_len: usize) -> %usize {
return @divTrunc(encoded_len, 4) * 3;
}
if (in_buf_len > 1) {
dest[dest_index] = ascii6[source[src_index + 0]] << 2 |
ascii6[source[src_index + 1]] >> 4;
dest_index += 1;
/// Invalid characters that are not ignored result in error.InvalidCharacter.
/// Invalid padding results in error.InvalidPadding.
/// Decoding more data than can fit in dest results in error.OutputTooSmall. See also ::calcSizeUpperBound.
/// Returns the number of bytes writen to dest.
pub fn decode(decoder_with_ignore: &const Base64DecoderWithIgnore, dest: []u8, source: []const u8) -> %usize {
const decoder = &const decoder_with_ignore.decoder;
var src_cursor: usize = 0;
var dest_cursor: usize = 0;
while (true) {
// get the next 4 chars, if available
var next_4_chars: [4]u8 = undefined;
var available_chars: usize = 0;
var pad_char_count: usize = 0;
while (available_chars < 4 and src_cursor < source.len) {
var c = source[src_cursor];
src_cursor += 1;
if (decoder.char_in_alphabet[c]) {
// normal char
next_4_chars[available_chars] = c;
available_chars += 1;
} else if (decoder_with_ignore.char_is_ignored[c]) {
// we're told to skip this one
continue;
} else if (c == decoder.pad_char) {
// the padding has begun. count the pad chars.
pad_char_count += 1;
while (src_cursor < source.len) {
c = source[src_cursor];
src_cursor += 1;
if (c == decoder.pad_char) {
pad_char_count += 1;
if (pad_char_count > 2) return error.InvalidCharacter;
} else if (decoder_with_ignore.char_is_ignored[c]) {
// we can even ignore chars during the padding
continue;
} else return error.InvalidCharacter;
}
break;
} else return error.InvalidCharacter;
}
switch (available_chars) {
4 => {
// common case
if (dest_cursor + 3 > dest.len) return error.OutputTooSmall;
assert(pad_char_count == 0);
dest[dest_cursor + 0] = decoder.char_to_index[next_4_chars[0]] << 2 |
decoder.char_to_index[next_4_chars[1]] >> 4;
dest[dest_cursor + 1] = decoder.char_to_index[next_4_chars[1]] << 4 |
decoder.char_to_index[next_4_chars[2]] >> 2;
dest[dest_cursor + 2] = decoder.char_to_index[next_4_chars[2]] << 6 |
decoder.char_to_index[next_4_chars[3]];
dest_cursor += 3;
continue;
},
3 => {
if (dest_cursor + 2 > dest.len) return error.OutputTooSmall;
if (pad_char_count != 1) return error.InvalidPadding;
dest[dest_cursor + 0] = decoder.char_to_index[next_4_chars[0]] << 2 |
decoder.char_to_index[next_4_chars[1]] >> 4;
dest[dest_cursor + 1] = decoder.char_to_index[next_4_chars[1]] << 4 |
decoder.char_to_index[next_4_chars[2]] >> 2;
if (decoder.char_to_index[next_4_chars[2]] << 6 != 0) return error.InvalidPadding;
dest_cursor += 2;
break;
},
2 => {
if (dest_cursor + 1 > dest.len) return error.OutputTooSmall;
if (pad_char_count != 2) return error.InvalidPadding;
dest[dest_cursor + 0] = decoder.char_to_index[next_4_chars[0]] << 2 |
decoder.char_to_index[next_4_chars[1]] >> 4;
if (decoder.char_to_index[next_4_chars[1]] << 4 != 0) return error.InvalidPadding;
dest_cursor += 1;
break;
},
1 => {
return error.InvalidPadding;
},
0 => {
if (pad_char_count != 0) return error.InvalidPadding;
break;
},
else => unreachable,
}
}
assert(src_cursor == source.len);
return dest_cursor;
}
if (in_buf_len > 2) {
dest[dest_index] = ascii6[source[src_index + 1]] << 4 |
ascii6[source[src_index + 2]] >> 2;
dest_index += 1;
}
if (in_buf_len > 3) {
dest[dest_index] = ascii6[source[src_index + 2]] << 6 |
ascii6[source[src_index + 3]];
dest_index += 1;
};
pub const standard_decoder_unsafe = Base64DecoderUnsafe.init(standard_alphabet_chars, standard_pad_char);
pub const Base64DecoderUnsafe = struct {
/// e.g. 'A' => 0.
/// undefined for any value not in the 64 alphabet chars.
char_to_index: [256]u8,
pad_char: u8,
pub fn init(alphabet_chars: []const u8, pad_char: u8) -> Base64DecoderUnsafe {
assert(alphabet_chars.len == 64);
var result = Base64DecoderUnsafe {
.char_to_index = undefined,
.pad_char = pad_char,
};
for (alphabet_chars) |c, i| {
assert(c != pad_char);
result.char_to_index[c] = u8(i);
}
return result;
}
return dest[0..dest_index];
}
pub fn calcEncodedSize(source_len: usize) -> usize {
return (((source_len * 4) / 3 + 3) / 4) * 4;
}
/// Computes the upper bound of the decoded size based only on the encoded length.
/// To compute the exact decoded size, see ::calcExactDecodedSize
pub fn calcMaxDecodedSize(encoded_len: usize) -> usize {
return @divExact(encoded_len * 3, 4);
}
/// Computes the number of decoded bytes there will be. This function must
/// be given the encoded buffer because there might be padding
/// bytes at the end ('=' in the standard alphabet)
pub fn calcExactDecodedSize(encoded: []const u8) -> usize {
return calcExactDecodedSizeWithAlphabet(encoded, standard_alphabet);
}
pub fn calcExactDecodedSizeWithAlphabet(encoded: []const u8, alphabet: []const u8) -> usize {
assert(alphabet.len == 65);
return calcExactDecodedSizeWithPadChar(encoded, alphabet[64]);
}
pub fn calcExactDecodedSizeWithPadChar(encoded: []const u8, pad_char: u8) -> usize {
var buf_len = encoded.len;
while (buf_len > 0 and encoded[buf_len - 1] == pad_char) {
buf_len -= 1;
/// The source buffer must be valid.
pub fn calcSize(decoder: &const Base64DecoderUnsafe, source: []const u8) -> usize {
return calcDecodedSizeExactUnsafe(source, decoder.pad_char);
}
return (buf_len * 3) / 4;
/// dest.len must be what you get from ::calcDecodedSizeExactUnsafe.
/// invalid characters or padding will result in undefined values.
pub fn decode(decoder: &const Base64DecoderUnsafe, dest: []u8, source: []const u8) {
assert(dest.len == decoder.calcSize(source));
var src_index: usize = 0;
var dest_index: usize = 0;
var in_buf_len: usize = source.len;
while (in_buf_len > 0 and source[in_buf_len - 1] == decoder.pad_char) {
in_buf_len -= 1;
}
while (in_buf_len > 4) {
dest[dest_index] = decoder.char_to_index[source[src_index + 0]] << 2 |
decoder.char_to_index[source[src_index + 1]] >> 4;
dest_index += 1;
dest[dest_index] = decoder.char_to_index[source[src_index + 1]] << 4 |
decoder.char_to_index[source[src_index + 2]] >> 2;
dest_index += 1;
dest[dest_index] = decoder.char_to_index[source[src_index + 2]] << 6 |
decoder.char_to_index[source[src_index + 3]];
dest_index += 1;
src_index += 4;
in_buf_len -= 4;
}
if (in_buf_len > 1) {
dest[dest_index] = decoder.char_to_index[source[src_index + 0]] << 2 |
decoder.char_to_index[source[src_index + 1]] >> 4;
dest_index += 1;
}
if (in_buf_len > 2) {
dest[dest_index] = decoder.char_to_index[source[src_index + 1]] << 4 |
decoder.char_to_index[source[src_index + 2]] >> 2;
dest_index += 1;
}
if (in_buf_len > 3) {
dest[dest_index] = decoder.char_to_index[source[src_index + 2]] << 6 |
decoder.char_to_index[source[src_index + 3]];
dest_index += 1;
}
}
};
fn calcDecodedSizeExactUnsafe(source: []const u8, pad_char: u8) -> usize {
if (source.len == 0) return 0;
var result = @divExact(source.len, 4) * 3;
if (source[source.len - 1] == pad_char) {
result -= 1;
if (source[source.len - 2] == pad_char) {
result -= 1;
}
}
return result;
}
test "base64" {
testBase64();
comptime testBase64();
@setEvalBranchQuota(5000);
%%testBase64();
comptime %%testBase64();
}
fn testBase64() {
testBase64Case("", "");
testBase64Case("f", "Zg==");
testBase64Case("fo", "Zm8=");
testBase64Case("foo", "Zm9v");
testBase64Case("foob", "Zm9vYg==");
testBase64Case("fooba", "Zm9vYmE=");
testBase64Case("foobar", "Zm9vYmFy");
fn testBase64() -> %void {
%return testAllApis("", "");
%return testAllApis("f", "Zg==");
%return testAllApis("fo", "Zm8=");
%return testAllApis("foo", "Zm9v");
%return testAllApis("foob", "Zm9vYg==");
%return testAllApis("fooba", "Zm9vYmE=");
%return testAllApis("foobar", "Zm9vYmFy");
%return testDecodeIgnoreSpace("", " ");
%return testDecodeIgnoreSpace("f", "Z g= =");
%return testDecodeIgnoreSpace("fo", " Zm8=");
%return testDecodeIgnoreSpace("foo", "Zm9v ");
%return testDecodeIgnoreSpace("foob", "Zm9vYg = = ");
%return testDecodeIgnoreSpace("fooba", "Zm9v YmE=");
%return testDecodeIgnoreSpace("foobar", " Z m 9 v Y m F y ");
// test getting some api errors
%return testError("A", error.InvalidPadding);
%return testError("AA", error.InvalidPadding);
%return testError("AAA", error.InvalidPadding);
%return testError("A..A", error.InvalidCharacter);
%return testError("AA=A", error.InvalidCharacter);
%return testError("AA/=", error.InvalidPadding);
%return testError("A/==", error.InvalidPadding);
%return testError("A===", error.InvalidCharacter);
%return testError("====", error.InvalidCharacter);
%return testOutputTooSmallError("AA==");
%return testOutputTooSmallError("AAA=");
%return testOutputTooSmallError("AAAA");
%return testOutputTooSmallError("AAAAAA==");
}
fn testBase64Case(expected_decoded: []const u8, expected_encoded: []const u8) {
const calculated_decoded_len = calcExactDecodedSize(expected_encoded);
assert(calculated_decoded_len == expected_decoded.len);
fn testAllApis(expected_decoded: []const u8, expected_encoded: []const u8) -> %void {
// Base64Encoder
{
var buffer: [0x100]u8 = undefined;
var encoded = buffer[0..Base64Encoder.calcSize(expected_decoded.len)];
standard_encoder.encode(encoded, expected_decoded);
assert(mem.eql(u8, encoded, expected_encoded));
}
const calculated_encoded_len = calcEncodedSize(expected_decoded.len);
assert(calculated_encoded_len == expected_encoded.len);
// Base64Decoder
{
var buffer: [0x100]u8 = undefined;
var decoded = buffer[0..%return standard_decoder.calcSize(expected_encoded)];
%return standard_decoder.decode(decoded, expected_encoded);
assert(mem.eql(u8, decoded, expected_decoded));
}
var buf: [100]u8 = undefined;
// Base64DecoderWithIgnore
{
const standard_decoder_ignore_nothing = Base64DecoderWithIgnore.init(
standard_alphabet_chars, standard_pad_char, "");
var buffer: [0x100]u8 = undefined;
var decoded = buffer[0..%return Base64DecoderWithIgnore.calcSizeUpperBound(expected_encoded.len)];
var written = %return standard_decoder_ignore_nothing.decode(decoded, expected_encoded);
assert(written <= decoded.len);
assert(mem.eql(u8, decoded[0..written], expected_decoded));
}
const actual_decoded = decode(buf[0..], expected_encoded);
assert(actual_decoded.len == expected_decoded.len);
assert(mem.eql(u8, expected_decoded, actual_decoded));
const actual_encoded = encode(buf[0..], expected_decoded);
assert(actual_encoded.len == expected_encoded.len);
assert(mem.eql(u8, expected_encoded, actual_encoded));
// Base64DecoderUnsafe
{
var buffer: [0x100]u8 = undefined;
var decoded = buffer[0..standard_decoder_unsafe.calcSize(expected_encoded)];
standard_decoder_unsafe.decode(decoded, expected_encoded);
assert(mem.eql(u8, decoded, expected_decoded));
}
}
fn testDecodeIgnoreSpace(expected_decoded: []const u8, encoded: []const u8) -> %void {
const standard_decoder_ignore_space = Base64DecoderWithIgnore.init(
standard_alphabet_chars, standard_pad_char, " ");
var buffer: [0x100]u8 = undefined;
var decoded = buffer[0..%return Base64DecoderWithIgnore.calcSizeUpperBound(encoded.len)];
var written = %return standard_decoder_ignore_space.decode(decoded, encoded);
assert(mem.eql(u8, decoded[0..written], expected_decoded));
}
error ExpectedError;
fn testError(encoded: []const u8, expected_err: error) -> %void {
const standard_decoder_ignore_space = Base64DecoderWithIgnore.init(
standard_alphabet_chars, standard_pad_char, " ");
var buffer: [0x100]u8 = undefined;
if (standard_decoder.calcSize(encoded)) |decoded_size| {
var decoded = buffer[0..decoded_size];
if (standard_decoder.decode(decoded, encoded)) |_| {
return error.ExpectedError;
} else |err| if (err != expected_err) return err;
} else |err| if (err != expected_err) return err;
if (standard_decoder_ignore_space.decode(buffer[0..], encoded)) |_| {
return error.ExpectedError;
} else |err| if (err != expected_err) return err;
}
fn testOutputTooSmallError(encoded: []const u8) -> %void {
const standard_decoder_ignore_space = Base64DecoderWithIgnore.init(
standard_alphabet_chars, standard_pad_char, " ");
var buffer: [0x100]u8 = undefined;
var decoded = buffer[0..calcDecodedSizeExactUnsafe(encoded, standard_pad_char) - 1];
if (standard_decoder_ignore_space.decode(decoded, encoded)) |_| {
return error.ExpectedError;
} else |err| if (err != error.OutputTooSmall) return err;
}

14
std/os/index.zig
View File

@ -31,6 +31,8 @@ pub const windowsWaitSingle = windows_util.windowsWaitSingle;
pub const windowsWrite = windows_util.windowsWrite;
pub const windowsIsCygwinPty = windows_util.windowsIsCygwinPty;
pub const windowsOpen = windows_util.windowsOpen;
pub const windowsLoadDll = windows_util.windowsLoadDll;
pub const windowsUnloadDll = windows_util.windowsUnloadDll;
pub const createWindowsEnvBlock = windows_util.createWindowsEnvBlock;
pub const FileHandle = if (is_windows) windows.HANDLE else i32;
@ -620,7 +622,9 @@ pub fn symLinkPosix(allocator: &Allocator, existing_path: []const u8, new_path:
}
// here we replace the standard +/ with -_ so that it can be used in a file name
const b64_fs_alphabet = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_=";
const b64_fs_encoder = base64.Base64Encoder.init(
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_",
base64.standard_pad_char);
pub fn atomicSymLink(allocator: &Allocator, existing_path: []const u8, new_path: []const u8) -> %void {
if (symLink(allocator, existing_path, new_path)) {
@ -632,12 +636,12 @@ pub fn atomicSymLink(allocator: &Allocator, existing_path: []const u8, new_path:
}
var rand_buf: [12]u8 = undefined;
const tmp_path = %return allocator.alloc(u8, new_path.len + base64.calcEncodedSize(rand_buf.len));
const tmp_path = %return allocator.alloc(u8, new_path.len + base64.Base64Encoder.calcSize(rand_buf.len));
defer allocator.free(tmp_path);
mem.copy(u8, tmp_path[0..], new_path);
while (true) {
%return getRandomBytes(rand_buf[0..]);
_ = base64.encodeWithAlphabet(tmp_path[new_path.len..], rand_buf, b64_fs_alphabet);
b64_fs_encoder.encode(tmp_path[new_path.len..], rand_buf);
if (symLink(allocator, existing_path, tmp_path)) {
return rename(allocator, tmp_path, new_path);
} else |err| {
@ -715,11 +719,11 @@ pub fn copyFile(allocator: &Allocator, source_path: []const u8, dest_path: []con
/// Guaranteed to be atomic.
pub fn copyFileMode(allocator: &Allocator, source_path: []const u8, dest_path: []const u8, mode: usize) -> %void {
var rand_buf: [12]u8 = undefined;
const tmp_path = %return allocator.alloc(u8, dest_path.len + base64.calcEncodedSize(rand_buf.len));
const tmp_path = %return allocator.alloc(u8, dest_path.len + base64.Base64Encoder.calcSize(rand_buf.len));
defer allocator.free(tmp_path);
mem.copy(u8, tmp_path[0..], dest_path);
%return getRandomBytes(rand_buf[0..]);
_ = base64.encodeWithAlphabet(tmp_path[dest_path.len..], rand_buf, b64_fs_alphabet);
b64_fs_encoder.encode(tmp_path[dest_path.len..], rand_buf);
var out_file = %return io.File.openWriteMode(tmp_path, mode, allocator);
defer out_file.close();

6
std/os/windows/index.zig
View File

@ -84,6 +84,11 @@ pub extern "kernel32" stdcallcc fn WriteFile(in_hFile: HANDLE, in_lpBuffer: &con
in_nNumberOfBytesToWrite: DWORD, out_lpNumberOfBytesWritten: ?&DWORD,
in_out_lpOverlapped: ?&OVERLAPPED) -> BOOL;
//TODO: call unicode versions instead of relying on ANSI code page
pub extern "kernel32" stdcallcc fn LoadLibraryA(lpLibFileName: LPCSTR) -> ?HMODULE;
pub extern "kernel32" stdcallcc fn FreeLibrary(hModule: HMODULE) -> BOOL;
pub extern "user32" stdcallcc fn MessageBoxA(hWnd: ?HANDLE, lpText: ?LPCTSTR, lpCaption: ?LPCTSTR, uType: UINT) -> c_int;
pub const PROV_RSA_FULL = 1;
@ -97,6 +102,7 @@ pub const FLOAT = f32;
pub const HANDLE = &c_void;
pub const HCRYPTPROV = ULONG_PTR;
pub const HINSTANCE = &@OpaqueType();
pub const HMODULE = &@OpaqueType();
pub const INT = c_int;
pub const LPBYTE = &BYTE;
pub const LPCH = &CHAR;

23
std/os/windows/util.zig
View File

@ -4,6 +4,7 @@ const windows = std.os.windows;
const assert = std.debug.assert;
const mem = std.mem;
const BufMap = std.BufMap;
const cstr = std.cstr;
error WaitAbandoned;
error WaitTimeOut;
@ -149,3 +150,25 @@ pub fn createWindowsEnvBlock(allocator: &mem.Allocator, env_map: &const BufMap)
result[i] = 0;
return result;
}
error DllNotFound;
pub fn windowsLoadDll(allocator: &mem.Allocator, dll_path: []const u8) -> %windows.HMODULE {
const padded_buff = %return cstr.addNullByte(allocator, dll_path);
defer allocator.free(padded_buff);
return windows.LoadLibraryA(padded_buff.ptr) ?? error.DllNotFound;
}
pub fn windowsUnloadDll(hModule: windows.HMODULE) {
assert(windows.FreeLibrary(hModule)!= 0);
}
test "InvalidDll" {
const DllName = "asdf.dll";
const allocator = std.debug.global_allocator;
const handle = os.windowsLoadDll(allocator, DllName) %% |err| {
assert(err == error.DllNotFound);
return;
};
}

44
test/cases/union.zig
View File

@ -31,3 +31,47 @@ test "unions embedded in aggregate types" {
else => unreachable,
}
}
const Foo = union {
float: f64,
int: i32,
};
test "basic unions" {
var foo = Foo { .int = 1 };
assert(foo.int == 1);
foo = Foo {.float = 12.34};
assert(foo.float == 12.34);
}
test "init union with runtime value" {
var foo: Foo = undefined;
setFloat(&foo, 12.34);
assert(foo.float == 12.34);
setInt(&foo, 42);
assert(foo.int == 42);
}
fn setFloat(foo: &Foo, x: f64) {
*foo = Foo { .float = x };
}
fn setInt(foo: &Foo, x: i32) {
*foo = Foo { .int = x };
}
const FooExtern = extern union {
float: f64,
int: i32,
};
test "basic extern unions" {
var foo = FooExtern { .int = 1 };
assert(foo.int == 1);
foo.float = 12.34;
assert(foo.float == 12.34);
}

25
test/compile_errors.zig
View File

@ -389,8 +389,8 @@ pub fn addCases(cases: &tests.CompileErrorContext) {
\\ const y = a.bar;
\\}
,
".tmp_source.zig:4:6: error: no member named 'foo' in 'A'",
".tmp_source.zig:5:16: error: no member named 'bar' in 'A'");
".tmp_source.zig:4:6: error: no member named 'foo' in struct 'A'",
".tmp_source.zig:5:16: error: no member named 'bar' in struct 'A'");
cases.add("redefinition of struct",
\\const A = struct { x : i32, };
@ -454,7 +454,7 @@ pub fn addCases(cases: &tests.CompileErrorContext) {
\\ .foo = 42,
\\ };
\\}
, ".tmp_source.zig:10:9: error: no member named 'foo' in 'A'");
, ".tmp_source.zig:10:9: error: no member named 'foo' in struct 'A'");
cases.add("invalid break expression",
\\export fn f() {
@ -2343,4 +2343,23 @@ pub fn addCases(cases: &tests.CompileErrorContext) {
\\pub extern fn foo(format: &const u8, ...);
,
".tmp_source.zig:2:9: error: expected type '&const u8', found '[5]u8'");
cases.add("constant inside comptime function has compile error",
\\const ContextAllocator = MemoryPool(usize);
\\
\\pub fn MemoryPool(comptime T: type) -> type {
\\ const free_list_t = @compileError("aoeu");
\\
\\ struct {
\\ free_list: free_list_t,
\\ }
\\}
\\
\\export fn entry() {
\\ var allocator: ContextAllocator = undefined;
\\}
,
".tmp_source.zig:4:25: error: aoeu",
".tmp_source.zig:1:36: note: called from here",
".tmp_source.zig:12:20: note: referenced here");
}

20
test/debug_safety.zig
View File

@ -260,4 +260,24 @@ pub fn addCases(cases: &tests.CompareOutputContext) {
\\ return int_slice[0];
\\}
);
cases.addDebugSafety("bad union field access",
\\pub fn panic(message: []const u8) -> noreturn {
\\ @import("std").os.exit(126);
\\}
\\
\\const Foo = union {
\\ float: f32,
\\ int: u32,
\\};
\\
\\pub fn main() -> %void {
\\ var f = Foo { .int = 42 };
\\ bar(&f);
\\}
\\
\\fn bar(f: &Foo) {
\\ f.float = 12.34;
\\}
);
}

48
test/tests.zig
View File

@ -18,7 +18,7 @@ const build_examples = @import("build_examples.zig");
const compile_errors = @import("compile_errors.zig");
const assemble_and_link = @import("assemble_and_link.zig");
const debug_safety = @import("debug_safety.zig");
const parsec = @import("parsec.zig");
const translate_c = @import("translate_c.zig");
const TestTarget = struct {
os: builtin.Os,
@ -123,16 +123,16 @@ pub fn addAssembleAndLinkTests(b: &build.Builder, test_filter: ?[]const u8) -> &
return cases.step;
}
pub fn addParseCTests(b: &build.Builder, test_filter: ?[]const u8) -> &build.Step {
const cases = %%b.allocator.create(ParseCContext);
*cases = ParseCContext {
pub fn addTranslateCTests(b: &build.Builder, test_filter: ?[]const u8) -> &build.Step {
const cases = %%b.allocator.create(TranslateCContext);
*cases = TranslateCContext {
.b = b,
.step = b.step("test-parsec", "Run the C header file parsing tests"),
.step = b.step("test-translate-c", "Run the C header file parsing tests"),
.test_index = 0,
.test_filter = test_filter,
};
parsec.addCases(cases);
translate_c.addCases(cases);
return cases.step;
}
@ -770,7 +770,7 @@ pub const BuildExamplesContext = struct {
}
};
pub const ParseCContext = struct {
pub const TranslateCContext = struct {
b: &build.Builder,
step: &build.Step,
test_index: usize,
@ -799,17 +799,17 @@ pub const ParseCContext = struct {
}
};
const ParseCCmpOutputStep = struct {
const TranslateCCmpOutputStep = struct {
step: build.Step,
context: &ParseCContext,
context: &TranslateCContext,
name: []const u8,
test_index: usize,
case: &const TestCase,
pub fn create(context: &ParseCContext, name: []const u8, case: &const TestCase) -> &ParseCCmpOutputStep {
pub fn create(context: &TranslateCContext, name: []const u8, case: &const TestCase) -> &TranslateCCmpOutputStep {
const allocator = context.b.allocator;
const ptr = %%allocator.create(ParseCCmpOutputStep);
*ptr = ParseCCmpOutputStep {
const ptr = %%allocator.create(TranslateCCmpOutputStep);
*ptr = TranslateCCmpOutputStep {
.step = build.Step.init("ParseCCmpOutput", allocator, make),
.context = context,
.name = name,
@ -821,7 +821,7 @@ pub const ParseCContext = struct {
}
fn make(step: &build.Step) -> %void {
const self = @fieldParentPtr(ParseCCmpOutputStep, "step", step);
const self = @fieldParentPtr(TranslateCCmpOutputStep, "step", step);
const b = self.context.b;
const root_src = %%os.path.join(b.allocator, b.cache_root, self.case.sources.items[0].filename);
@ -829,7 +829,7 @@ pub const ParseCContext = struct {
var zig_args = ArrayList([]const u8).init(b.allocator);
%%zig_args.append(b.zig_exe);
%%zig_args.append("parsec");
%%zig_args.append("translate-c");
%%zig_args.append(b.pathFromRoot(root_src));
warn("Test {}/{} {}...", self.test_index+1, self.context.test_index, self.name);
@ -882,7 +882,7 @@ pub const ParseCContext = struct {
if (stderr.len != 0 and !self.case.allow_warnings) {
warn(
\\====== parsec emitted warnings: ============
\\====== translate-c emitted warnings: =======
\\{}
\\============================================
\\
@ -914,7 +914,7 @@ pub const ParseCContext = struct {
warn("\n");
}
pub fn create(self: &ParseCContext, allow_warnings: bool, filename: []const u8, name: []const u8,
pub fn create(self: &TranslateCContext, allow_warnings: bool, filename: []const u8, name: []const u8,
source: []const u8, expected_lines: ...) -> &TestCase
{
const tc = %%self.b.allocator.create(TestCase);
@ -932,37 +932,37 @@ pub const ParseCContext = struct {
return tc;
}
pub fn add(self: &ParseCContext, name: []const u8, source: []const u8, expected_lines: ...) {
pub fn add(self: &TranslateCContext, name: []const u8, source: []const u8, expected_lines: ...) {
const tc = self.create(false, "source.h", name, source, expected_lines);
self.addCase(tc);
}
pub fn addC(self: &ParseCContext, name: []const u8, source: []const u8, expected_lines: ...) {
pub fn addC(self: &TranslateCContext, name: []const u8, source: []const u8, expected_lines: ...) {
const tc = self.create(false, "source.c", name, source, expected_lines);
self.addCase(tc);
}
pub fn addAllowWarnings(self: &ParseCContext, name: []const u8, source: []const u8, expected_lines: ...) {
pub fn addAllowWarnings(self: &TranslateCContext, name: []const u8, source: []const u8, expected_lines: ...) {
const tc = self.create(true, "source.h", name, source, expected_lines);
self.addCase(tc);
}
pub fn addCase(self: &ParseCContext, case: &const TestCase) {
pub fn addCase(self: &TranslateCContext, case: &const TestCase) {
const b = self.b;
const annotated_case_name = %%fmt.allocPrint(self.b.allocator, "parsec {}", case.name);
const annotated_case_name = %%fmt.allocPrint(self.b.allocator, "translate-c {}", case.name);
if (self.test_filter) |filter| {
if (mem.indexOf(u8, annotated_case_name, filter) == null)
return;
}
const parsec_and_cmp = ParseCCmpOutputStep.create(self, annotated_case_name, case);
self.step.dependOn(&parsec_and_cmp.step);
const translate_c_and_cmp = TranslateCCmpOutputStep.create(self, annotated_case_name, case);
self.step.dependOn(&translate_c_and_cmp.step);
for (case.sources.toSliceConst()) |src_file| {
const expanded_src_path = %%os.path.join(b.allocator, b.cache_root, src_file.filename);
const write_src = b.addWriteFile(expanded_src_path, src_file.source);
parsec_and_cmp.step.dependOn(&write_src.step);
translate_c_and_cmp.step.dependOn(&write_src.step);
}
}
};

342
test/parsec.zig → test/translate_c.zig
View File

@ -1,6 +1,6 @@
const tests = @import("tests.zig");
pub fn addCases(cases: &tests.ParseCContext) {
pub fn addCases(cases: &tests.TranslateCContext) {
cases.addAllowWarnings("simple data types",
\\#include <stdint.h>
\\int foo(char a, unsigned char b, signed char c);
@ -677,12 +677,12 @@ pub fn addCases(cases: &tests.ParseCContext) {
\\ };
\\ a >>= @import("std").math.Log2Int(c_int)({
\\ const _ref = &a;
\\ (*_ref) = c_int(c_int(*_ref) >> @import("std").math.Log2Int(c_int)(1));
\\ (*_ref) = ((*_ref) >> @import("std").math.Log2Int(c_int)(1));
\\ *_ref
\\ });
\\ a <<= @import("std").math.Log2Int(c_int)({
\\ const _ref = &a;
\\ (*_ref) = c_int(c_int(*_ref) << @import("std").math.Log2Int(c_int)(1));
\\ (*_ref) = ((*_ref) << @import("std").math.Log2Int(c_int)(1));
\\ *_ref
\\ });
\\}
@ -735,12 +735,12 @@ pub fn addCases(cases: &tests.ParseCContext) {
\\ };
\\ a >>= @import("std").math.Log2Int(c_uint)({
\\ const _ref = &a;
\\ (*_ref) = c_uint(c_uint(*_ref) >> @import("std").math.Log2Int(c_uint)(1));
\\ (*_ref) = ((*_ref) >> @import("std").math.Log2Int(c_uint)(1));
\\ *_ref
\\ });
\\ a <<= @import("std").math.Log2Int(c_uint)({
\\ const _ref = &a;
\\ (*_ref) = c_uint(c_uint(*_ref) << @import("std").math.Log2Int(c_uint)(1));
\\ (*_ref) = ((*_ref) << @import("std").math.Log2Int(c_uint)(1));
\\ *_ref
\\ });
\\}
@ -805,6 +805,50 @@ pub fn addCases(cases: &tests.ParseCContext) {
\\}
);
cases.addC("pre increment/decrement",
\\void foo(void) {
\\ int i = 0;
\\ unsigned u = 0;
\\ ++i;
\\ --i;
\\ ++u;
\\ --u;
\\ i = ++i;
\\ i = --i;
\\ u = ++u;
\\ u = --u;
\\}
,
\\export fn foo() {
\\ var i: c_int = 0;
\\ var u: c_uint = c_uint(0);
\\ i += 1;
\\ i -= 1;
\\ u +%= 1;
\\ u -%= 1;
\\ i = {
\\ const _ref = &i;
\\ (*_ref) += 1;
\\ *_ref
\\ };
\\ i = {
\\ const _ref = &i;
\\ (*_ref) -= 1;
\\ *_ref
\\ };
\\ u = {
\\ const _ref = &u;
\\ (*_ref) +%= 1;
\\ *_ref
\\ };
\\ u = {
\\ const _ref = &u;
\\ (*_ref) -%= 1;
\\ *_ref
\\ };
\\}
);
cases.addC("do loop",
\\void foo(void) {
\\ int a = 2;
@ -834,17 +878,21 @@ pub fn addCases(cases: &tests.ParseCContext) {
cases.addC("deref function pointer",
\\void foo(void) {}
\\void baz(void) {}
\\void bar(void) {
\\ void(*f)(void) = foo;
\\ f();
\\ (*(f))();
\\ baz();
\\}
,
\\export fn foo() {}
\\export fn baz() {}
\\export fn bar() {
\\ var f: ?extern fn() = foo;
\\ (??f)();
\\ (??f)();
\\ baz();
\\}
);
@ -857,15 +905,277 @@ pub fn addCases(cases: &tests.ParseCContext) {
\\ (*(??x)) = 1;
\\}
);
cases.add("simple union",
\\union Foo {
\\ int x;
\\ double y;
\\};
,
\\pub const union_Foo = extern union {
\\ x: c_int,
\\ y: f64,
\\};
,
\\pub const Foo = union_Foo;
);
cases.add("address of operator",
\\int foo(void) {
\\ int x = 1234;
\\ int *ptr = &x;
\\ return *ptr;
\\}
,
\\pub fn foo() -> c_int {
\\ var x: c_int = 1234;
\\ var ptr: ?&c_int = &x;
\\ return *(??ptr);
\\}
);
cases.add("string literal",
\\const char *foo(void) {
\\ return "bar";
\\}
,
\\pub fn foo() -> ?&const u8 {
\\ return c"bar";
\\}
);
cases.add("return void",
\\void foo(void) {
\\ return;
\\}
,
\\pub fn foo() {
\\ return;
\\}
);
cases.add("for loop",
\\void foo(void) {
\\ for (int i = 0; i < 10; i += 1) { }
\\}
,
\\pub fn foo() {
\\ {
\\ var i: c_int = 0;
\\ while (i < 10) : (i += 1) {};
\\ };
\\}
);
cases.add("empty for loop",
\\void foo(void) {
\\ for (;;) { }
\\}
,
\\pub fn foo() {
\\ while (true) {};
\\}
);
cases.add("break statement",
\\void foo(void) {
\\ for (;;) {
\\ break;
\\ }
\\}
,
\\pub fn foo() {
\\ while (true) {
\\ break;
\\ };
\\}
);
cases.add("continue statement",
\\void foo(void) {
\\ for (;;) {
\\ continue;
\\ }
\\}
,
\\pub fn foo() {
\\ while (true) {
\\ continue;
\\ };
\\}
);
cases.add("switch statement",
\\int foo(int x) {
\\ switch (x) {
\\ case 1:
\\ x += 1;
\\ case 2:
\\ break;
\\ case 3:
\\ case 4:
\\ return x + 1;
\\ default:
\\ return 10;
\\ }
\\ return x + 13;
\\}
,
\\fn foo(_arg_x: c_int) -> c_int {
\\ var x = _arg_x;
\\ {
\\ switch (x) {
\\ 1 => goto case_0,
\\ 2 => goto case_1,
\\ 3 => goto case_2,
\\ 4 => goto case_3,
\\ else => goto default,
\\ };
\\ case_0:
\\ x += 1;
\\ case_1:
\\ goto end;
\\ case_2:
\\ case_3:
\\ return x + 1;
\\ default:
\\ return 10;
\\ goto end;
\\ end:
\\ };
\\ return x + 13;
\\}
);
cases.add("macros with field targets",
\\typedef unsigned int GLbitfield;
\\typedef void (*PFNGLCLEARPROC) (GLbitfield mask);
\\typedef void(*OpenGLProc)(void);
\\union OpenGLProcs {
\\ OpenGLProc ptr[1];
\\ struct {
\\ PFNGLCLEARPROC Clear;
\\ } gl;
\\};
\\extern union OpenGLProcs glProcs;
\\#define glClearUnion glProcs.gl.Clear
\\#define glClearPFN PFNGLCLEARPROC
,
\\pub const GLbitfield = c_uint;
,
\\pub const PFNGLCLEARPROC = ?extern fn(GLbitfield);
,
\\pub const OpenGLProc = ?extern fn();
,
\\pub const union_OpenGLProcs = extern union {
\\ ptr: [1]OpenGLProc,
\\ gl: extern struct {
\\ Clear: PFNGLCLEARPROC,
\\ },
\\};
,
\\pub extern var glProcs: union_OpenGLProcs;
,
\\pub const glClearPFN = PFNGLCLEARPROC;
,
\\pub inline fn glClearUnion(arg0: GLbitfield) {
\\ (??glProcs.gl.Clear)(arg0)
\\}
,
\\pub const OpenGLProcs = union_OpenGLProcs;
);
cases.add("switch statement with no default",
\\int foo(int x) {
\\ switch (x) {
\\ case 1:
\\ x += 1;
\\ case 2:
\\ break;
\\ case 3:
\\ case 4:
\\ return x + 1;
\\ }
\\ return x + 13;
\\}
,
\\fn foo(_arg_x: c_int) -> c_int {
\\ var x = _arg_x;
\\ {
\\ switch (x) {
\\ 1 => goto case_0,
\\ 2 => goto case_1,
\\ 3 => goto case_2,
\\ 4 => goto case_3,
\\ else => goto end,
\\ };
\\ case_0:
\\ x += 1;
\\ case_1:
\\ goto end;
\\ case_2:
\\ case_3:
\\ return x + 1;
\\ goto end;
\\ end:
\\ };
\\ return x + 13;
\\}
);
cases.add("variable name shadowing",
\\int foo(void) {
\\ int x = 1;
\\ {
\\ int x = 2;
\\ x += 1;
\\ }
\\ return x;
\\}
,
\\pub fn foo() -> c_int {
\\ var x: c_int = 1;
\\ {
\\ var x_0: c_int = 2;
\\ x_0 += 1;
\\ };
\\ return x;
\\}
);
cases.add("pointer casting",
\\float *ptrcast(int *a) {
\\ return (float *)a;
\\}
,
\\fn ptrcast(a: ?&c_int) -> ?&f32 {
\\ return @ptrCast(?&f32, a);
\\}
);
cases.add("bin not",
\\int foo(int x) {
\\ return ~x;
\\}
,
\\pub fn foo(x: c_int) -> c_int {
\\ return ~x;
\\}
);
cases.add("primitive types included in defined symbols",
\\int foo(int u32) {
\\ return u32;
\\}
,
\\pub fn foo(u32_0: c_int) -> c_int {
\\ return u32_0;
\\}
);
cases.add("const ptr initializer",
\\static const char *v0 = "0.0.0";
,
\\pub var v0: ?&const u8 = c"0.0.0";
);
}
// TODO
//float *ptrcast(int *a) {
// return (float *)a;
//}
// should translate to
// fn ptrcast(a: ?&c_int) -> ?&f32 {
// return @ptrCast(?&f32, a);
// }