MiNT (Atari) port

[mmuman]

Summary

This adds support for the Atari MiNT OS (not the GNU/Linux distro).

Changes

• a lot of detection modules,
• some code required to build on the OS which does not support all the fancy new POSIX things.

Screenshots

Checklist

• I have tested my changes locally.

[mmuman]

damn there's no config.h …

[CarterLi]

Thanks for working on this port — it is clearly a serious effort, and I appreciate the amount of research and implementation work that went into it.

That said, I am not in favor of merging this PR in its current form.

My main concern is not that MiNT is niche by itself, but that this adds a new long-term maintenance surface for a platform that the project is unlikely to have active maintainers, regular users, or any CI coverage for. In practice, that means:

• future changes to shared structs, module interfaces, or common helpers can silently break MiNT code,
• nobody may notice those regressions for a long time,
• and the burden eventually shifts to core maintainers who do not use or understand the platform.

This PR also does not stay fully isolated in platform-specific files. It modifies a fair amount of shared/common code to accommodate MiNT. That raises the maintenance cost for everyone, not just MiNT users. For a platform with no verification pipeline and uncertain long-term usage, that tradeoff seems hard to justify.

I am also uncomfortable with some of the compatibility techniques used here, especially changes like:

• replacing static_assert with an empty macro,
• redefining or bypassing low-level APIs just to force compilation,
• and more generally introducing compatibility paths that appear to hide unsupported behavior rather than model it cleanly.

Even if these choices are understandable for getting an initial port running, they are difficult to accept in core shared code because they reduce confidence in correctness and make the codebase harder to reason about.

So my current view is:

• As a proof of concept / external port: this is interesting and valuable.
• As mainline support in fastfetch: I do not think it is ready, and I am not convinced the long-term maintenance cost is worth it.

I would be more open to reconsidering this if the port could be reworked to:

1. keep almost all changes isolated to MiNT-specific files,
2. remove the more aggressive compatibility hacks from shared headers/common code,
3. document a realistic maintenance story,
4. and ideally provide at least some reproducible way to validate builds going forward.

Without that, I think merging this would likely create dead code over time rather than sustainable platform support.

[mmuman]

I'll be busy the rest of the month, but I'll look into that. Some thing will be hard to do though, unless adding a statat syscall to MiNT… But maybe the usecase is simple enough it can be emulated. Some things might be better off with feature checks than ifdeffing on the OS.

Maybe some function should be moved to something the platform can override instead of the ifdeffery… Or platform-specific includes?

It should be possible to add proper CI since the cross tools are packaged for Debian, never did that on github so it's an occasion I guess. Might need to make a few gigs of space here though… Might even get ARAnyM with a headless console 🤔

[codacy-production]

Not up to standards ⛔

🔴 Issues 10 medium

Alerts:
⚠ 10 issues (≤ 0 issues of at least minor severity)

Results:
10 new issues

Category	Results
ErrorProne	10 medium

View in Codacy

🟢 Metrics 417 complexity · 10 duplication

Metric	Results
Complexity	417
Duplication	10

View in Codacy

AI Reviewer: first review requested successfully. AI can make mistakes. Always validate suggestions.

_{TIP This summary will be updated as you push new changes.}

[codacy-production]

Pull Request Overview

This PR introduces Atari MiNT support but contains several critical bugs that prevent correct operation on the target platform. Most notably, the logic for identifying hardware via system cookies is inverted, and the package caching mechanism is fundamentally broken. Directory recursion is effectively disabled because the fallback for missing POSIX directory APIs always identifies entries as non-directories.

From a security and stability perspective, there is a memory safety violation in the string processing module that will cause data corruption on Big Endian systems. Additionally, the toolchain configuration contains hardcoded absolute paths that will break builds outside of specific environments. These issues must be addressed before the port can be considered functional.

About this PR

• The use of stubs for 'openat' and 'fstatat' without proper fallbacks (like 'stat') leads to systemic failures in directory processing. A more robust compatibility layer is needed.
• The Atari toolchain files contain hardcoded absolute paths which limit portability. These should be replaced with relative paths or environment variables.

Test suggestions

• Verify BIOS version and date detection on various Atari TOS/MiNT versions
• Verify Board model identification via the _MCH cookie across ST, STE, Falcon, and ARAnyM
• Verify CPU type and frequency detection for 68000, 68030, 68060 (CT60), and ColdFire
• Verify that ffTimeGetTick correctly converts _hz_200 ticks to milliseconds on MiNT
• Test application behavior when relative file operations (openat) are stubbed to fail
• Improve unit test coverage for complex logic in src/common/impl/FFPlatform_unix.c
• Improve unit test coverage for complex logic in src/common/impl/io_unix.c
• Improve unit test coverage for complex logic in src/common/impl/processing_linux.c

Prompt proposal for missing tests

Consider implementing these tests if applicable:
1. Verify BIOS version and date detection on various Atari TOS/MiNT versions
2. Verify Board model identification via the _MCH cookie across ST, STE, Falcon, and ARAnyM
3. Verify CPU type and frequency detection for 68000, 68030, 68060 (CT60), and ColdFire
4. Test application behavior when relative file operations (openat) are stubbed to fail
5. Improve unit test coverage for complex logic in src/common/impl/FFPlatform_unix.c
6. Improve unit test coverage for complex logic in src/common/impl/io_unix.c
7. Improve unit test coverage for complex logic in src/common/impl/processing_linux.c

_{TIP Improve review quality by adding custom instructions
TIP How was this review? Give us feedback}

[codacy-production]

_{🔴 HIGH RISK}

Recursive directory traversal is broken for MiNT. The variable isDir is never set to true because the fstatat and S_ISDIR checks are excluded. As this file is high-complexity and lacks coverage, this failure is high risk. Use stat() or lstat() as a fallback when fstatat is unavailable to allow directory recursion.

[codacy-production]

_{🔴 HIGH RISK}

The logic is inverted. Getcookie returns non-zero (1) on success. The current implementation will return an error message when the cookie is actually found.