gh-131591: Handle includes for iOS in remote_debugging.c

[pablogsal]

• Issue: Implement PEP 768 – Safe external debugger interface for CPython #131591

[pablogsal]

@freakboy3742 can you advise if this is the best way to handle this? The error was:

HEAD is now at 943cc1431eb gh-131591: Implement PEP 768 (#131937)
Switched to and reset branch 'main'

configure: WARNING: no system libmpdecimal found; falling back to bundled libmpdecimal (deprecated and scheduled for removal in Python 3.15)
configure: WARNING: pkg-config is missing. Some dependencies may not be detected correctly.

configure: WARNING: no system libmpdecimal found; falling back to bundled libmpdecimal (deprecated and scheduled for removal in Python 3.15)
configure: WARNING: pkg-config is missing. Some dependencies may not be detected correctly.

../../Python/remote_debugging.c:30:14: fatal error: 'libproc.h' file not found
   30 | #    include <libproc.h>
      |              ^~~~~~~~~~~
1 error generated.
make: *** [Python/remote_debugging.o] Error 1

find: build: No such file or directory
find: build: No such file or directory
find: build: No such file or directory
find: build: No such file or directory
make: [clean-retain-profile] Error 1 (ignored)

[pablogsal]

!buildbot simulator

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @pablogsal for commit 3ec2116 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F132050%2Fmerge

The command will test the builders whose names match following regular expression: simulator

The builders matched are:

• iOS ARM64 Simulator PR

[pablogsal]

!buildbot simulator

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @pablogsal for commit 1e29ec6 🤖

Results will be shown at:

https://buildbot.python.org/all/#/grid?branch=refs%2Fpull%2F132050%2Fmerge

The command will test the builders whose names match following regular expression: simulator

The builders matched are:

• iOS ARM64 Simulator PR

[freakboy3742]

This will break on older macOS versions (essentially any macOS version that predates the existence of iPhone) because TARGET_OS_OSX doesn't exist. However, on all versions of iOS, TARGET_OS_OSX will exist with a value of 0. So - my suggested approach here would be:

Suggested change

	#if defined(__APPLE__) && TARGET_OS_OSX
	# include <libproc.h>
	#if defined(__APPLE__)
	# include <TargetConditionals.h>
	# if !defined(TARGET_OS_OSX) || TARGET_OS_OSX

I can see there's a couple of other places (L99 of this file; L352 of pycore_eval.h) where an analogous change would be required.

I guess the other option would be to modify the code that sets TARGET_OS_OSX from:

#  if !defined(TARGET_OS_OSX)
 #     define TARGET_OS_OSX 1
 #  endif

to

#  if !defined(TARGET_OS_OSX) && !(defined(TARGET_OS_IPHONE) && TARGET_OS_IPHONE)
#     define TARGET_OS_OSX 1
#  endif

This works because TARGET_OS_IPHONE is any Apple mobile device (iOS, tvOS, watchOS, visionOS); TARGET_OS_IOS is specifically an iPhone/iPad. However, futzing with TARGET_OS_OSX seems like asking for trouble, so the explicit approach would be my preferred option.

[pablogsal]

I think you may be missing the context that this file is importing this header where we are doing the dance:

cpython/Include/internal/pycore_ceval.h

Lines 350 to 360 in 0dbaeb9

	#ifndef Py_SUPPORTS_REMOTE_DEBUG
	#if defined(__APPLE__)
	# if !defined(TARGET_OS_OSX)
	// Older macOS SDKs do not define TARGET_OS_OSX
	# define TARGET_OS_OSX 1
	# endif
	#endif
	#if ((defined(__APPLE__) && TARGET_OS_OSX) || defined(MS_WINDOWS) || (defined(__linux__) && HAVE_PROCESS_VM_READV))
	# define Py_SUPPORTS_REMOTE_DEBUG 1
	#endif
	#endif

[freakboy3742]

Yes - I did miss that import would be in effect. Given that imported definition, the form you've got in the PR at present should work - but it still makes me a bit twitchy because it's redefining an Apple-provided constant that has otherwise well defined behavior.

This is essentially what bit me here - the logic didn't make sense because I didn't notice the import where the behavior of the Apple-provided constant was being redefined). We've also had examples in the past where someone adds/removes an import of an otherwise unrelated file, and suddenly things break because that file has redefined basic platform constants (or has removed an import that was making those constants work in a predictable way).

If we're going to go down the path of redefining the constant, it might be worth looking at a top-level replacement for TargetConditionals.h that uses those base definitions and ensures they're consistently defined (so we don't need to do the defined(TARGET_OS_OSX) && TARGET_OS_OSX dance everywhere) - but that's a much bigger set of changes, and then comes with the overhead of making sure that everyone remembers to use the updated definitions, rather than importing TargetConditionals.h.

So - call my preference for using the definitions-as-provided a "strong opinion, weakly held". With the indentation nitpick flagged in pycore_ceval.h, I can live with this as-is.

[bedevere-app]

When you're done making the requested changes, leave the comment: I have made the requested changes; please review again.

[pablogsal]

@freakboy3742 If you prefer we can use the if !defined(TARGET_OS_OSX) && !(defined(TARGET_OS_IPHONE) && TARGET_OS_IPHONE) version. Would you prefer that instead of the current approach? I selected this based on your code in _testexternalinspection.c

[freakboy3742]

I selected this based on your code in _testexternalinspection.c

... huh. Well now I've been hoist with my own petard.

I've gone on a bit of a code dive to see how big a change reverting #define TARGET_OS_OSX 1 would be, and the more I look, the more the "redefining" approach seems the pragmatic choice. There's lots of "bare" uses of TARGET_OS_OSX that are already impacted by the existing definitions, and none of those are made easier to read by requiring an explicit defined(TARGET_OS_OSX) as a prefix. And, if someone forgets that the defined(TARGET_OS_OSX) is required, that's an automatic breakage for old macOS versions - so having the redefine in place is a good safety catch.

So - lets go with the redefine-based approach you've got here.