gh-117886: platform_triplet.c: check if TARGET_OS_* defined

[jmroot]

Older macOS SDKs don't define any of these, and some clang versions will error if you use them without first checking if they are defined.

• Issue: platform_triplet.c uses possibly undefined TARGET_OS_* macros #117886

[freakboy3742]

I'm intrigued by how old the SDK needs to be to not have these symbols. The use of the OSX name means the symbol should have been introduced any later than 2015, as that was when OS X 10.11 (the last "OS X" version) was released.

Regardless, these changes make sense to me. However, there are also uses of the same symbols in:

• Modules/_testexternalinspection.c (seems to be addressed in _testexternalinspection.c doesn't check for undefined TARGET_OS_OSX #117891?)
• Python/marshal.c
• Objects/mimalloc/prim/unix/prim.c (these definitions predate the PEP 730 work, and use symbols that were deprecated in iOS 15)

Not sure if we want to consolidate all the TARGET_OS fixes into one PR, or if we have a third PR addressing the marshal.c and prim.c versions.

[erlend-aasland]

I'd be fine with one PR.

[ned-deily]

For context, @jmroot is reporting these issues as a maintainer of the MacPorts project which is well-known for providing best-effort support of (much) older versions of Mac hardware and macOS releases.

Since most of the problematic uses were introduced as part of #114099 "Add support for iOS" feature for the upcoming 3.13 release which is still in its alpha stage, my preference would be to treat those as feature fixes and associate any PRs with that issue rather than creating new issues. But I don't have a strong preference about that nor whether to consolidate PRs.

However, I do have a stronger preference about any fix for the pre-existing problematic use in Objects/mimalloc/prim/unix/prim.c since, I believe, that code was essentially vendored into Python from an existing project and thus should be tracked separately and either checked for an existing upstream fix or reported upstream.

[freakboy3742]

Since most of the problematic uses were introduced as part of #114099 "Add support for iOS" feature for the upcoming 3.13 release which is still in its alpha stage, my preference would be to treat those as feature fixes and associate any PRs with that issue rather than creating new issues. But I don't have a strong preference about that nor whether to consolidate PRs.

I've submitted #118073 as a consolidation of the three non-prim.c changes, tagged against the iOS PR.

However, I do have a stronger preference about any fix for the pre-existing problematic use in Objects/mimalloc/prim/unix/prim.c since, I believe, that code was essentially vendored into Python from an existing project and thus should be tracked separately and either checked for an existing upstream fix or reported upstream.

I've opened #118072 to describe the prim.c problem; I've logged the details of my initial investigation there.

[ned-deily]

Thanks, @jmroot, for reporting the issue(s) and for providing the initial PRs. And, thanks, @freakboy3742, for the follow-ups. Since #118073 incorporates this change, we can close this PR and its issue.