diff --git a/LANGUAGE.md b/LANGUAGE.md index 1468dc3..248d389 100644 --- a/LANGUAGE.md +++ b/LANGUAGE.md @@ -289,7 +289,7 @@ exactly once. Bare functions named `memcopy` or `memset` remain ordinary user fu - `std/meta` reflection records plus `EnumFieldStruct(E, Field, default ?Field)`, implemented with `struct_type!`; it produces a record with one field per native enum member in declaration order, where outer `null` means no field default - `std/io` explicit `Io` capabilities, provider-bound `Reader`/`Writer` handles, existing-file open/close operations, allocation-free `write_all`, and comptime-expanded writer-first `print`; formatting supports natural `{}`, byte `{s}`, decimal `{d}`, integer `{b}` / `{o}` / `{x}` / `{X}`, byte-character `{c}`, scientific float `{e}`, recursively scalar-backed distinct values, and `{{` / `}}`, with malformed formats and incompatible tuple fields rejected at comptime - entry points are either `main func() ...` or `main func(init process.Init) ...`; their success channel is `void`, `i32`, or `int` and may have an error channel; an unhandled entry error exits with status 1. `std/process.Init` contains exactly `io io.Io` and `args Args`, while the system provider remains hidden inside `std/io` -- `std/process.Args` contains exactly `values [][*;0]u8`, borrowing native `argv` and its `argc` count without allocation or copying; index 0 is the executable name when present. `argument_count(args Args) usize` returns the count, and `argument(args Args, index usize) ?[;0]u8` scans only the requested argument and returns a borrowed sentinel slice (including empty arguments), or null out of range; borrowed views remain valid while the supplied storage lives +- `std/process.Args` contains exactly `values [][;0]u8`; use `args.values.len` and direct indexing. Native startup scans each argument once into heap-owned slice descriptors while borrowing, not copying, the strings; index 0 is the executable name when present, and empty arguments remain empty slices. The descriptors and their borrowed views are valid until `main` returns; the native wrapper frees the descriptors on both success and error - `std/debug.print` is an allocation-free, failure-ignoring stderr escape hatch independent of `process.Init` - `std/testing` supplies fallible `expect`, expected-first `expect_equal`, and exact compile-time `expect_type`; direct calls through an alias of exactly `@std/testing` receive compiler-injected source locations @@ -300,7 +300,7 @@ exactly once. Bare functions named `memcopy` or `memset` remain ordinary user fu - lazy semantic checking of demanded function specializations - static, eager runtime, mutable runtime, and deferred problematic globals with cycle diagnostics - demand-driven LLVM declarations for referenced foreign functions -- root `main` may be parameterless or accept canonical `@std/process Init`; the generated C entry point accepts native `argc`/`argv`, obtains the hidden system I/O provider, and constructs the init value using computed record layout. Both `Init` and `Args` must have their canonical auto-layout record shapes +- root `main` may be parameterless or accept canonical `@std/process Init`; the generated C entry point accepts native `argc`/`argv`, obtains the hidden system I/O provider, and constructs the init value using computed record layout. Both `Init` and `Args` must have their canonical auto-layout record shapes. Descriptor allocation is size-checked, zero-count safe, and exits with status 2 on overflow or allocation failure; parameterless `main` does not prepare arguments - `brolang test [root]` reuses `build.bro`, discovers only explicit test-import edges, skips the application entry point, and runs tests sequentially while continuing after assertion failures - replaceable dynamically loaded libclang C-import backend diff --git a/README.md b/README.md index b779fbf..73378c7 100644 --- a/README.md +++ b/README.md @@ -24,14 +24,15 @@ main func(init process.Init) void { } ``` -`Init` contains `io io.Io` and `args process.Args`. `Args.values` is a borrowed -`[][*;0]u8` view of native `argv`, with its length taken from `argc`: it includes -the executable name at index 0 when present, without copying strings or the -pointer array. `process.argument_count(init.args)` returns that count, and -`process.argument(init.args, index)` returns a borrowed `?[;0]u8` (null when out -of range). The helper scans only the requested zero-terminated argument; empty -arguments remain empty slices. The slices are valid while their supplied -argument storage lives. +`Init` contains `io io.Io` and `args process.Args`. `Args.values` has type +`[][;0]u8`: use `init.args.values.len` for the count and +`init.args.values[index]` for a borrowed, zero-terminated argument slice. +Index 0 is the executable name when present, and empty arguments are empty +slices. Startup scans each native argument once and allocates only the slice +descriptors; argument strings are never copied. The native entry wrapper owns +the descriptors until `main` returns, freeing them on both success and error. +An empty native argument list is supported; allocation overflow or failure exits +with status 2. Parameterless `main` does not prepare unused arguments. Bodyless `c_func` declarations bind exact external symbols and require concrete types. C primitives use atomic target-dependent names and remain semantically diff --git a/compiler/checker/checker.odin b/compiler/checker/checker.odin index 9c18369..4042718 100644 --- a/compiler/checker/checker.odin +++ b/compiler/checker/checker.odin @@ -1764,15 +1764,15 @@ configure_entry_point :: proc(checker: ^Checker) { args_item, args_ok := types.node(&checker.module.types, args_type) args_fields := types.fields_for(&checker.module.types, args_type) values_name := symbol.intern(checker.symbols, "values") - argument_pointer := types.pointer(&checker.module.types, types.U8, false, true, true, 0) - argument_values := types.slice(&checker.module.types, argument_pointer, false) + argument_slice := types.slice(&checker.module.types, types.U8, false, true, 0) + argument_values := types.slice(&checker.module.types, argument_slice, false) if !args_ok || args_item.kind != .Struct || args_item.tuple || args_item.c_layout || len(args_fields) != 1 || args_fields[0].name != u32(values_name) || !types.equal(types.resolve_alias(args_fields[0].type, &checker.module.types), argument_values) { checker.template_diagnostics[main_template] = source.add( checker.diagnostics, main.span, - "@std/process Args must be an auto-layout record containing exactly 'values [][*;0]u8'", + "@std/process Args must be an auto-layout record containing exactly 'values [][;0]u8'", ) return } diff --git a/compiler/ir/ir.odin b/compiler/ir/ir.odin index 8869582..6b0107d 100644 --- a/compiler/ir/ir.odin +++ b/compiler/ir/ir.odin @@ -86,6 +86,9 @@ Opcode :: enum u8 { Store, Fill, Slice, + // Private native-entry operations: own descriptors, borrow argument strings. + Prepare_Args, + Free_Args, Length, Slice_Ptr, Extract, diff --git a/compiler/llvm/llvm.odin b/compiler/llvm/llvm.odin index b6f6652..d6af7cf 100644 --- a/compiler/llvm/llvm.odin +++ b/compiler/llvm/llvm.odin @@ -20,6 +20,7 @@ Emitter :: struct { builder: strings.Builder, messages: [dynamic]Trap_Message, allocator: mem.Allocator, + prepares_args: bool, } C_Record_ABI_Kind :: enum u8 { @@ -284,7 +285,7 @@ valid_value :: proc( } switch instructions[value_id].op { case .Param, .Const, .Poison, .String, .Aggregate, .Null, .Optional_Some, - .Load_Global, .Function_Address, .Address_Of, .Load, .Union_Tag, .Slice, .Length, .Slice_Ptr, + .Load_Global, .Function_Address, .Address_Of, .Load, .Union_Tag, .Slice, .Prepare_Args, .Length, .Slice_Ptr, .Fallible_Error, .Extract, .Select, .Unwrap, .Optional_Is_Some, .Optional_Value, .Orelse, .Widen, .Sum_Widen, .Sum_Project, .C_Coerce, .C_Vararg_Promote, .Retype, .Scalar_Cast, .Pointer_Cast, .Const_Cast, .Weaken_Pointer, .Weaken_Slice, .Decay_Array_Pointer, @@ -295,7 +296,7 @@ valid_value :: proc( .Shift_Left, .Shift_Right, .Shift_Left_Saturating, .Compare, .Call: return true case .Address_Global, .Alloca, .Index_Address, .Field_Address, .Orelse_Begin, - .Store, .Fill, .Mem_Copy, .Mem_Set, .Trap, .Label, .Br, .Cond_Br, .Return, .Return_Void: + .Store, .Fill, .Free_Args, .Mem_Copy, .Mem_Set, .Trap, .Label, .Br, .Cond_Br, .Return, .Return_Void: return false } return false @@ -1545,6 +1546,16 @@ emit_instruction_stream :: proc( write_operand(&emitter.builder, instructions, instruction.b, instruction.type, &emitter.module.types) fmt.sbprintf(&emitter.builder, ", ptr %%memset_element%d\n", instruction_index) fmt.sbprintf(&emitter.builder, " %%memset_next%d = add i64 %%memset_index%d, 1\n store i64 %%memset_next%d, ptr %%memset_index_slot%d\n br label %%memset_loop%d\nmemset_done%d:\n", instruction_index, instruction_index, instruction_index, instruction_index, instruction_index, instruction_index) + case .Prepare_Args: + fmt.sbprintf(&emitter.builder, " %%v%d = call %s @bro.prepare_args(i32 ", instruction_index, llvm_type(instruction.type, &emitter.module.types)) + write_operand(&emitter.builder, instructions, instruction.a, types.C_INT, &emitter.module.types) + strings.write_string(&emitter.builder, ", ptr ") + write_operand(&emitter.builder, instructions, instruction.b, instructions[instruction.b].type, &emitter.module.types) + strings.write_string(&emitter.builder, ")\n") + case .Free_Args: + fmt.sbprintf(&emitter.builder, " %%args_data%d = extractvalue %s ", instruction_index, llvm_type(instructions[instruction.a].type, &emitter.module.types)) + write_operand(&emitter.builder, instructions, instruction.a, instructions[instruction.a].type, &emitter.module.types) + fmt.sbprintf(&emitter.builder, ", 0\n call void @free(ptr %%args_data%d)\n", instruction_index) case .Slice: if !valid_instruction(instructions, instruction.a) { emit_recovery_value(emitter, instruction_index, instruction, "invalid slice container") @@ -2746,6 +2757,11 @@ emit_functions :: proc(emitter: ^Emitter) { if function.implementation == .Declaration && function.link_name == "write" { continue } + if emitter.prepares_args && function.implementation == .Declaration && + (function.link_name == "malloc" || function.link_name == "free" || + function.link_name == "strlen" || function.link_name == "exit") { + continue + } if function.implementation == .Declaration { duplicate := false for previous in emitter.module.functions[:function_index] { @@ -2880,6 +2896,59 @@ emit_declarations :: proc(emitter: ^Emitter) { &emitter.builder, "\ndefine internal void @bro.trap(ptr %message, i64 %length) noreturn {\nentry:\n %written = call i64 @write(i32 2, ptr %message, i64 %length)\n call void @llvm.trap()\n unreachable\n}\n\n", ) + if emitter.prepares_args { + emit_argument_runtime(emitter) + } +} + +// Only native entry wrappers own these descriptors; argument bytes stay borrowed. +emit_argument_runtime :: proc(emitter: ^Emitter) { + strings.write_string(&emitter.builder, `declare ptr @malloc(i64) +declare void @free(ptr) +declare i64 @strlen(ptr) +declare void @exit(i32) noreturn + +define internal { ptr, i64 } @bro.prepare_args(i32 %argc, ptr %argv) { +entry: + %negative = icmp slt i32 %argc, 0 + br i1 %negative, label %failure, label %check_empty +check_empty: + %count = zext i32 %argc to i64 + %empty = icmp eq i64 %count, 0 + br i1 %empty, label %empty_result, label %check_size +check_size: + %allocation = call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %count, i64 16) + %size = extractvalue { i64, i1 } %allocation, 0 + %overflow = extractvalue { i64, i1 } %allocation, 1 + br i1 %overflow, label %failure, label %allocate +allocate: + %data = call ptr @malloc(i64 %size) + %failed = icmp eq ptr %data, null + br i1 %failed, label %failure, label %loop +loop: + %index = phi i64 [ 0, %allocate ], [ %next, %loop ] + %native = getelementptr ptr, ptr %argv, i64 %index + %string = load ptr, ptr %native + %length = call i64 @strlen(ptr %string) + %descriptor = getelementptr { ptr, i64 }, ptr %data, i64 %index + %with_pointer = insertvalue { ptr, i64 } poison, ptr %string, 0 + %argument = insertvalue { ptr, i64 } %with_pointer, i64 %length, 1 + store { ptr, i64 } %argument, ptr %descriptor + %next = add i64 %index, 1 + %more = icmp ult i64 %next, %count + br i1 %more, label %loop, label %done +done: + %result_pointer = insertvalue { ptr, i64 } poison, ptr %data, 0 + %result = insertvalue { ptr, i64 } %result_pointer, i64 %count, 1 + ret { ptr, i64 } %result +empty_result: + ret { ptr, i64 } zeroinitializer +failure: + call void @exit(i32 2) + unreachable +} + +`) } emit :: proc( @@ -2895,6 +2964,14 @@ emit :: proc( builder=strings.builder_make(allocator), allocator=allocator, } + for function in module.functions { + for instruction in function.instructions { + if instruction.op == .Prepare_Args { + emitter.prepares_args = true + break + } + } + } emitter.messages.allocator = allocator defer { for message in emitter.messages { diff --git a/compiler/lower/lower.odin b/compiler/lower/lower.odin index 1e193cc..87bc078 100644 --- a/compiler/lower/lower.odin +++ b/compiler/lower/lower.odin @@ -1784,6 +1784,7 @@ append_injected_main :: proc(module: ^ir.Module, hir_module: ^hir.Module, alloca instructions.allocator = allocator param_types: []types.Type args: []ir.Instruction_Id + prepared_args := ir.INVALID_INSTRUCTION if param_index >= 0 { init_type := main_function.locals[param_index].type init_fields := types.fields_for(&module.types, init_type) @@ -1797,7 +1798,7 @@ append_injected_main :: proc(module: ^ir.Module, hir_module: ^hir.Module, alloca args_index := 1 - io_index args_type := types.resolve_alias(init_fields[args_index].type, &module.types) values_type := types.resolve_alias(types.fields_for(&module.types, args_type)[0].type, &module.types) - argument_pointer := types.child_type(values_type, &module.types) + argument_pointer := types.pointer(&module.types, types.U8, false, true, true, 0) argv_type := types.pointer(&module.types, argument_pointer, false, true) param_types = make([]types.Type, 2, allocator) param_types[0] = types.C_INT @@ -1809,23 +1810,14 @@ append_injected_main :: proc(module: ^ir.Module, hir_module: ^hir.Module, alloca diagnostic=source.INVALID_DIAGNOSTIC, }) } - count := ir.instruction_id(len(instructions)) + prepared_args = ir.instruction_id(len(instructions)) append(&instructions, ir.Instruction{ - op=.Scalar_Cast, type=types.USIZE, target=ir.INVALID_REF, - a=ir.Instruction_Id(0), b=ir.INVALID_INSTRUCTION, - diagnostic=source.INVALID_DIAGNOSTIC, - }) - slice_args := make([]ir.Instruction_Id, 2, allocator) - slice_args[0] = ir.INVALID_INSTRUCTION - slice_args[1] = count - values := ir.instruction_id(len(instructions)) - append(&instructions, ir.Instruction{ - op=.Slice, type=values_type, args=slice_args, target=ir.INVALID_REF, - a=ir.Instruction_Id(1), b=ir.INVALID_INSTRUCTION, + op=.Prepare_Args, type=values_type, target=ir.INVALID_REF, + a=ir.Instruction_Id(0), b=ir.Instruction_Id(1), diagnostic=source.INVALID_DIAGNOSTIC, }) values_args := make([]ir.Instruction_Id, 1, allocator) - values_args[0] = values + values_args[0] = prepared_args process_args := ir.instruction_id(len(instructions)) append(&instructions, ir.Instruction{ op=.Aggregate, type=args_type, args=values_args, target=ir.INVALID_REF, @@ -1867,6 +1859,13 @@ append_injected_main :: proc(module: ^ir.Module, hir_module: ^hir.Module, alloca b=ir.INVALID_INSTRUCTION, diagnostic=source.INVALID_DIAGNOSTIC, }) + if prepared_args != ir.INVALID_INSTRUCTION { + append(&instructions, ir.Instruction{ + op=.Free_Args, type=types.VOID, target=ir.INVALID_REF, + a=prepared_args, b=ir.INVALID_INSTRUCTION, + diagnostic=source.INVALID_DIAGNOSTIC, + }) + } if types.kind(main_function.result, &hir_module.types) == .Fallible { success := types.fallible_success(main_function.result, &hir_module.types) diff --git a/compiler_tests.odin b/compiler_tests.odin index 2867559..ed0c0c0 100644 --- a/compiler_tests.odin +++ b/compiler_tests.odin @@ -3534,10 +3534,28 @@ main func() int ! Error { return try value() } source=`process :: import "@std/process" Error :: enum { failed } fail func() void ! Error { return .failed } -main func(_ process.Init) void ! Error { try fail() } +main func(init process.Init) void ! Error { + if init.args.values.len != 3 { return } + if init.args.values[1].len != 0 or init.args.values[2].len != 9 { return } + if init.args.values[2][0] != 't' or init.args.values[2][8] != 's' { return } + try fail() +} `, exit_code=1, }, + { + name="process-success", + source=`process :: import "@std/process" +Error :: enum { failed } +main func(init process.Init) i32 ! Error { + if init.args.values.len != 3 { return .failed } + if init.args.values[1].len != 0 or init.args.values[2].len != 9 { return .failed } + if init.args.values[2][0] != 't' or init.args.values[2][8] != 's' { return .failed } + return 7 +} +`, + exit_code=7, + }, { name="inferred-main-try", source=`Error :: enum { failed } @@ -3578,7 +3596,12 @@ main func() void! { try dispatch(2) } source=`process :: import "@std/process" Error :: enum { failed } fail func() void ! Error { return .failed } -main func(_ process.Init) void! { try fail() } +main func(init process.Init) void! { + if init.args.values.len != 3 { return } + if init.args.values[1].len != 0 or init.args.values[2].len != 9 { return } + if init.args.values[2][0] != 't' or init.args.values[2][8] != 's' { return } + try fail() +} `, exit_code=1, }, @@ -3603,7 +3626,7 @@ main func(_ process.Init) void! { try fail() } ) testing.expect_value(t, status, 0) state, stdout, stderr, _ := os2.process_exec( - os2.Process_Desc{command=[]string{output}}, + os2.Process_Desc{command=[]string{output, "", "two words"}}, context.allocator, ) defer delete(stdout) @@ -3941,7 +3964,7 @@ milestone_37_validates_process_init_and_hidden_system_provider :: proc(t: ^testi system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*;0]u8 } +Args :: struct { values [][;0]u8 } Init :: struct { io io.Io, args Args } `, }, @@ -3950,7 +3973,7 @@ Init :: struct { io io.Io, args Args } @hide system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*;0]u8 } +Args :: struct { values [][;0]u8 } Init :: struct { io io.Io, args Args, extra i32 } `, }, @@ -3959,7 +3982,7 @@ Init :: struct { io io.Io, args Args, extra i32 } @hide system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*;0]u8 } +Args :: struct { values [][;0]u8 } Init :: struct { io io.Io } `, }, @@ -3968,7 +3991,7 @@ Init :: struct { io io.Io } @hide system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*]u8 } +Args :: struct { values [][]u8 } Init :: struct { io io.Io, args Args } `, }, @@ -3977,7 +4000,7 @@ Init :: struct { io io.Io, args Args } @hide system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*;0]u8, extra i32 } +Args :: struct { values [][;0]u8, extra i32 } Init :: struct { io io.Io, args Args } `, }, @@ -3986,7 +4009,7 @@ Init :: struct { io io.Io, args Args } @hide system func() Io { return Io {value = 0} } `, process_source=`io :: import "@std/io" -Args :: struct { values [][*;0]u8 } +Args :: struct { values [][;0]u8 } Init :: struct { io io.Io, args i32 } `, }, @@ -4030,7 +4053,7 @@ main func(init process.Init) void { _ = init } } @(test) -process_init_supplies_native_arguments_and_helpers_borrow_explicit_storage :: proc(t: ^testing.T) { +process_init_supplies_native_argument_slices :: proc(t: ^testing.T) { directory := "/tmp/brolang-test-process-args" main_path := "/tmp/brolang-test-process-args/main.bro" output := "/tmp/brolang-test-process-args/app" @@ -4038,43 +4061,23 @@ process_init_supplies_native_arguments_and_helpers_borrow_explicit_storage :: pr io :: import "@std/io" mem :: import "@std/mem" -supplied_args_work func() bool { - text [3;0]mut u8 := ['r', 'u', 'n'] - pointers [1][*;0]u8 :: [(&text).ptr] - args :: process.Args { values = pointers[..] } - if process.argument_count(args) != 1 { return false } - if process.argument(args, 0) |value| { - text[0] = 'f' - if !mem.eql(value, "fun") { return false } - } else { return false } - if process.argument(args, 1) |_| { return false } - empty :: process.Args { values = pointers[0..0] } - if process.argument_count(empty) != 0 { return false } - if process.argument(empty, 0) |_| { return false } - return true -} - main func(init process.Init) i32 { - count :: process.argument_count(init.args) - if count != 1 and count != 4 { return 1 } - if process.argument(init.args, 0) |executable| { - if !mem.eql(executable, "/tmp/brolang-test-process-args/app") { return 2 } - } else { return 3 } - if process.argument(init.args, count) |_| { return 4 } - if process.argument(init.args, maxval!(usize)) |_| { return 5 } + count :: init.args.values.len + if count != 1 and count != 4 and count != 16385 { return 1 } + if !mem.eql(init.args.values[0], "/tmp/brolang-test-process-args/app") { return 2 } if count == 4 { - if process.argument(init.args, 1) |operation| { - if !mem.eql(operation, "run") { return 6 } - } else { return 7 } - if process.argument(init.args, 2) |empty| { - if empty.len != 0 { return 8 } - } else { return 9 } - if process.argument(init.args, 3) |spaced| { - if !mem.eql(spaced, "two words") { return 10 } - } else { return 11 } + if !mem.eql(init.args.values[1], "run") { return 3 } + if init.args.values[2].len != 0 { return 4 } + if !mem.eql(init.args.values[3], "two words") { return 5 } } - if !supplied_args_work() { return 12 } - io.write_all(io.stdout(init.io), "args-ok\n") catch |_| { return 13 } + if count == 16385 { + index := usize(1) + while index < count { + if init.args.values[index].len != 0 { return 6 } + index += 1 + } + } + io.write_all(io.stdout(init.io), "args-ok\n") catch |_| { return 7 } return 0 } ` @@ -4085,9 +4088,16 @@ main func(init process.Init) i32 { testing.expect_value(t, compiler_core.compile_package( directory, output, nil, target.DEFAULT, cimport.Options{}, ".", ), 0) + large_command := make([]string, 16385) + defer delete(large_command) + large_command[0] = output + for index in 1..= argument_count(args) { - return null - } - pointer :: args.values[index] - length usize := 0 - while pointer[length] != 0 { - length += 1 - } - return pointer[..length;0] + values [][;0]u8 }