ffmpeg_{4,6,7}: Darwin clean‐ups and improvements

[emilazy]

This brings the functionality of our mpv package on macOS more in line with their official CI binaries while cleaning up a bunch of cruft that is no longer required after #346043. See the commit messages for details.

The flag removals are technically breaking changes past the freeze, although I highly doubt anyone was ever toggling them. If preferred, we could instead leave the flags around but warn about their impending removal when they’re set, and ditch them after branch‐off. I’d definitely rather not keep around pointless feature flags that don’t incur any additional dependencies in 25.05, though.

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 24.11 Release Notes (or backporting 23.11 and 24.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[emilazy]

Actually apparently we’re going based on anywhere‐on‐Earth time so the removals are totally okay.

[emilazy]

Retargeted to staging as this will of course cause a bunch of rebuilds.

[Atemu]

Diff LGTM. If you say this works the same on Darwin, that sounds good to me.

[emilazy]

It works better, at least for hardware‐accelerated decoding on new OS versions :)

It should behave the same way on old macOS releases but it’s always possible that upstream may have missed some availability checks. That would be a bug we can fix (and upstream the fix for) if it arises, but should be much less relevant once we bump the minimum supported version for 25.05 anyway.

[Atemu]

Speaking of which, we use SDK 15 here, wouldn't that already exclude macOS < 15 from running the binaries?

[emilazy]

No; the SDK controls what functionality you can access, and the minimum version (“deployment target”) controls what functionality is assumed to be available. With the new SDK pattern, apple-sdk_* determines the former and darwinMinVersionHook determines the latter. If you try to use APIs that are not present in your minimum version, you get a warning in C‐family languages and an error in Swift. Some packages are badly‐written and use autoconf‐style static checks to determine API availability, so we default to the SDK corresponding to our minimum supported version by default to reduce the likelihood of programs failing to start on old OS versions, but software that is correctly written for macOS will do conditional runtime availability checks for features they use that are not present in the minimum OS version they support.

In the case of FFmpeg, they even have checks for versions below our current x86_64-darwin baseline of 10.12:

emily@yuyuko ~/D/ffmpeg ((66d61918))> rg '@available|__builtin_available'
libavfilter/vf_yadif_videotoolbox.m
199:    if (@available(macOS 10.11, iOS 8.0, *)) {
279:    if (@available(macOS 10.11, iOS 8.0, *)) {
315:    if (@available(macOS 10.11, iOS 8.0, *)) {
383:    if (@available(macOS 10.11, iOS 8.0, *)) {

libavcodec/videotoolboxenc.c
1207:    if (__builtin_available(macOS 10.13, *)) {
1629:        if (__builtin_available(macOS 10.10, *)) {

libavfilter/metal/utils.m
33:    if (@available(macOS 10.15, iOS 11, tvOS 14.5, *)) {

libavutil/hwcontext_videotoolbox.c
442:        if (__builtin_available(macOS 10.11, iOS 9, *))
455:        if (__builtin_available(macOS 10.13, iOS 11, tvOS 11, watchOS 4, *))
468:        if (__builtin_available(macOS 10.11, iOS 9, *))
480:        if (__builtin_available(macOS 10.13, iOS 11, tvOS 11, watchOS 4, *))
494:        if (__builtin_available(macOS 10.13, iOS 11, *))
501:        if (__builtin_available(macOS 10.11, iOS 9, *))
511:        if (__builtin_available(macOS 10.12, iOS 10, *))
517:        if (__builtin_available(macOS 10.13, iOS 11, *))
527:        if (__builtin_available(macOS 10.13, iOS 11, tvOS 11, watchOS 4, *))
548:        if (__builtin_available(macOS 12.0, iOS 15.0, tvOS 15.0, *))
624:    if (__builtin_available(macOS 10.8, iOS 10, *)) {

libavcodec/videotoolbox.c
953:#if defined(MAC_OS_X_VERSION_10_9) && !TARGET_OS_IPHONE && (MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_X_VERSION_10_9) && AV_HAS_BUILTIN(__builtin_available)
955:        if (__builtin_available(macOS 10.9, *)) {
961:#if defined(MAC_OS_VERSION_11_0) && !TARGET_OS_IPHONE && (MAC_OS_X_VERSION_MAX_ALLOWED >= MAC_OS_VERSION_11_0) && AV_HAS_BUILTIN(__builtin_available)
962:    if (__builtin_available(macOS 11.0, *)) {

However, you did remind me that despite supporting deploying to old versions, FFmpeg doesn’t properly build with the ancient 10.12 SDK we currently default to on x86_64-darwin; they use a typedef that isn’t present and that we previously had to patch around. This is quite a common case; there is a lot of software like MoltenVK that supports quite old OS releases but expects a modern SDK to build with. I’ve now removed that patch, further reducing the Darwin‐related cruft 🎉

[nixos-discourse]

This pull request has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/title-the-darwin-sdks-have-been-updated/55295/1