ADR 0011: Product scope — Amiga GFA BASIC 3.5 to Linux x86-64 with SDL3
ADR 0011: Product scope — Amiga GFA BASIC 3.5 to Linux x86-64 with SDL3
- Status: Accepted
- Date: 2026-11-05
- Decision makers: Project maintainer
- Tracking: issues #3, #4, #52, and #53
Context
The project already selected Amiga GFA BASIC interpreter 3.5 as its normative source dialect (ADR 0009) and Linux x86-64 as its initial output target (ADR 0003). The earlier wording described these as initial choices and left room for multiple dialects, API profiles, and output targets. That open-ended wording creates unnecessary parallel work and obscures the intended product.
The product goal is a compiler for Amiga GFA BASIC 3.5 programs that produces native Linux x86-64 executables. Linux graphics, windowing, and input integration will use SDL3.
Decision
- The sole supported source dialect/version is Commodore Amiga GFA BASIC interpreter 3.5. This is the product target, not merely the first dialect in a planned matrix.
- The sole supported output target is Linux x86-64, using the GNU ABI and ELF 64-bit format (x86_64-unknown-linux-gnu).
- LLVM is the planned native code-generation backend.
- SDL3 is the selected Linux platform integration library for graphics, windows, and input. SDL3 calls must remain behind project-owned runtime interfaces in the Linux platform adapter.
- The Amiga GFA BASIC language core, Amiga API compatibility layer, generic runtime contract, LLVM backend, and Linux/SDL3 implementation remain separate architectural responsibilities.
- The compiler frontend and target-independent IR must not depend on LLVM or SDL3. The generated program must link or load SDL3 only when required by the runtime functionality it uses.
- Amiga 3.0/3.03 references remain historical evidence. Rules found only there must be confirmed against 3.5 before implementation or conformance claims.
- Atari ST dialects/API profiles, other Amiga interpreter/compiler versions, other BASIC dialects, other output operating systems/architectures, and cross-compilation are out of scope. Existing research artifacts may remain for historical context but do not create support commitments.
- Do not implement undocumented or ambiguous behavior by guessing. Unsupported source constructs and unavailable Linux capabilities must produce explicit diagnostics.
- This scope decision does not claim that the language, LLVM backend, runtime, SDL3 adapter, or end-to-end compiler is already implemented or conformant.
Alternatives considered
- Keep multiple dialects and platform profiles as near-term goals: rejected because the immediate product is specifically Amiga 3.5 to Linux and parallel targets dilute evidence and implementation effort.
- Implement Atari ST compatibility alongside Amiga: rejected; Atari evidence is not normative for Amiga and Atari support is outside the product scope.
- Use SDL3 directly from the compiler or language core: rejected because it would couple source semantics and IR to one platform library.
- Require SDL3 for every generated executable: rejected; non-graphical programs should not acquire an unnecessary graphics/input dependency.
- Add other output targets or cross-compilation immediately: rejected; the Linux x86-64 path must be completed and validated first.
Consequences
- Research and implementation issues must prioritize Amiga GFA BASIC 3.5 evidence only.
- The lexer, parser, semantic analysis, typed IR, LLVM backend, runtime ABI, and CLI should be designed to deliver the one defined product target.
- Linux/SDL3 work is the only platform integration work in scope.
- Atari ST research may remain archived, but Atari implementation tasks should be closed as not planned.
- The compatibility matrix and README must not imply a multi-dialect or multi-target product.
- New scope expansion requires a new explicit product decision.