refactor: optimizing the logic for retrieving free memory on iOS 13+

[94haox]

In iOS 13 and later, Apple has provided an API to obtain available memory.
link: https://developer.apple.com/documentation/os/3191911-os_proc_available_memory

[GLinnik21]

As far as I can understand, while os_proc_available_memory() is a more accurate measure for app-specific available memory, it represents a fundamentally different value than the VM stats approach:

1. Different metric: os_proc_available_memory() shows memory available to the process, while VM stats show system-wide free memory. These are not directly comparable.

2. Inconsistent reporting: Using os_proc_available_memory() for iOS 13+ and VM stats for earlier versions would create a discontinuity in reported values across iOS versions, potentially confusing users of the crash reporting system.

What do you think?

[94haox]

As far as I can understand, while os_proc_available_memory() is a more accurate measure for app-specific available memory, it represents a fundamentally different value than the VM stats approach:

1. Different metric: os_proc_available_memory() shows memory available to the process, while VM stats show system-wide free memory. These are not directly comparable.
2. Inconsistent reporting: Using os_proc_available_memory() for iOS 13+ and VM stats for earlier versions would create a discontinuity in reported values across iOS versions, potentially confusing users of the crash reporting system.

What do you think?

I believe your points are valid, but I have a different opinion:

Metric Standards: From the perspective of solving problems, for systems like iOS/tvOS/watchOS that only support single-process apps, the idle memory at the process level is more valuable compared to the system level idle memory (I guess this is also the reason why the API does not support macOS?);

Inconsistency in Reporting: Indeed, such problems may exist, but the current logic, I guess, is a compromise strategy when precise process memory information cannot be obtained (I'm not quite sure)? Compared to accurate information, the priority of report value continuity should be relatively low?

[GLinnik21]

Thanks for your perspective. The recent implementation in KSCrash seems to address our concerns effectively. You might want to check out these files for more details:

https://github.com/kstenerud/KSCrash/blob/962934327eb42ebee883eca7258df1fd03e2cf89/Sources/KSCrashRecording/KSCrashAppMemoryTracker.m
https://github.com/kstenerud/KSCrash/blob/962934327eb42ebee883eca7258df1fd03e2cf89/Sources/KSCrashRecording/KSCrashAppMemory.m

What do you think about this approach?

[94haox]

That's great, the feature you mentioned is very useful. It seems I haven't been paying attention to KSCrash for quite a while 😄.