ARchive old spec

2026-07-21 22:30:10 +02:00
parent 0380705e76
commit 184dec86ca
17 changed files with 550 additions and 191 deletions
@@ -0,0 +1,88 @@
+# text-assembler Specification
+
+## Purpose
+TBD - created by archiving change inject-and-assemble. Update Purpose after archive.
+## Requirements
+### Requirement: AsmBuilder assembles x86 instruction text to byte array
+
+`AsmBuilder.Assemble(string source)` parses x86 assembly text and returns the corresponding `byte[]` machine code.
+
+#### Scenario: single instruction
+- **WHEN** `AsmBuilder.Assemble("nop")` is called
+- **THEN** the result is `[0x90]`
+
+#### Scenario: multiple instructions
+- **WHEN** `AsmBuilder.Assemble("pushad\npopad")` is called
+- **THEN** the result is `[0x60, 0x61]`
+
+#### Scenario: instruction with immediate operand
+- **WHEN** `AsmBuilder.Assemble("mov eax, 1")` is called
+- **THEN** the result is `[0xB8, 0x01, 0x00, 0x00, 0x00]`
+
+### Requirement: AsmBuilder supports register operands
+
+Supported registers: `eax`, `ecx`, `edx`, `ebx`, `esp`, `ebp`, `esi`, `edi` (and 8-bit: `al`, `cl`, `dl`, `bl`, `ah`, `ch`, `dh`, `bh`).
+
+#### Scenario: register-to-register move
+- **WHEN** `AsmBuilder.Assemble("mov eax, ecx")` is called
+- **THEN** the result is `[0x89, 0xC8]` (mov eax, ecx encoding)
+
+#### Scenario: register encoding
+- **WHEN** registers are used in instructions
+- **THEN** each register maps to its correct 3-bit encoding (eax=0, ecx=1, edx=2, ebx=3, esp=4, ebp=5, esi=6, edi=7)
+
+### Requirement: AsmBuilder supports labels and jumps
+
+Labels are defined with `@name:` and referenced with `jmp @name` or `je @name`. Forward and backward references are resolved in a second pass.
+
+#### Scenario: forward jump
+- **WHEN** `AsmBuilder.Assemble("jmp @skip\nnop\n@skip:\nret")` is called
+- **THEN** the jump skips exactly over the `nop` (2 bytes) and lands on `ret`
+
+#### Scenario: backward jump
+- **WHEN** `AsmBuilder.Assemble("@loop:\nnop\njmp @loop")` is called
+- **THEN** the jump targets the earlier label correctly
+
+#### Scenario: multiple labels
+- **WHEN** multiple labels are used in one source
+- **THEN** each label resolves to its correct byte offset
+
+### Requirement: AsmBuilder SetPassLimit controls iteration
+
+`AsmBuilder.SetPassLimit(int limit)` sets the maximum number of assembly passes for label resolution. Default is 10. If the limit is exceeded before all labels resolve, `InvalidOperationException` is thrown.
+
+#### Scenario: default pass limit
+- **WHEN** no `SetPassLimit` is called
+- **THEN** the assembler uses 10 passes maximum
+
+#### Scenario: custom pass limit
+- **WHEN** `SetPassLimit(20)` is called
+- **THEN** the assembler uses 20 passes maximum
+
+#### Scenario: pass limit exceeded
+- **WHEN** forward references cannot resolve within the pass limit
+- **THEN** `InvalidOperationException` is thrown with label resolution details
+
+### Requirement: AsmBuilder reports clear errors
+
+Unknown instructions, missing operands, and invalid register names produce `ArgumentException` with the line number and offending text.
+
+#### Scenario: unknown instruction
+- **WHEN** `AsmBuilder.Assemble("xyzw")` is called
+- **THEN** `ArgumentException` is thrown mentioning line 1 and "xyzw"
+
+#### Scenario: missing operand
+- **WHEN** `AsmBuilder.Assemble("mov")` is called (no operands)
+- **THEN** `ArgumentException` is thrown mentioning missing operand
+
+### Requirement: AsmBuilder supported instruction set
+
+The following x86 instructions are supported:
+- **Data movement**: `mov`, `push`, `pop`, `pushad`, `popad`, `lea`
+- **Arithmetic**: `add`, `sub`, `inc`, `dec`, `xor`, `and`, `or`, `cmp`, `test`
+- **Control flow**: `jmp`, `je`, `jne`, `call`, `ret`, `nop`, `hlt`
+
+#### Scenario: all instructions produce valid bytes
+- **WHEN** each supported instruction is assembled individually
+- **THEN** it produces the correct x86 machine code encoding
+