Could not find enough documentation on below behavior, appreciate if someone could direct to the documentation of below behavior:Running App would terminate/restart when permission is altered in settings.Scenario:If App, using microphone is running and say is in screen#3 and microphone permission is granted.Now if user revokes the microphone permission by invoking Settings->Privacy->Microphone->App -> off and opens already running App, it starts from screen #1 instead of #3.What is the best way to prevent this behavior as microphone permission is not mandatory but convenient option for application and could wait for next time launch to apply permission change. Thinking to store state before going to background and reapply when relaunched. But is there a possibility to not to reset the app when permission changes at first place automatically, like Bluetooth on/off

Hi Team,I went through WWDC 2018 video on 405_measuring_performance_using_logging and would like to replace existing logging mechanism with os_log to gain its benifits with signposts to measure performance.But there is main blocker as listed below:1. Current logging mechanism provides easy way to share the App logs by click of button,without need of external systems like Mac/Windows/iTune and its specific to our app resulting in minimum file size in KBs.How could we achieve this by migrating to os_log, as it logs into a common bulky sysdiagnose file?1.Could we wrap os_log to log to a file in App's documents folder?Simple sample code will be helpful, as I could not find more info on wrapping os_log.2. Could I programatically access sysdiagnose with filters of "my catogery" and export into App's local folder to share as before?Thanks a ton in Advance!

Scenario: We have a situation where we have multple core ml models integrated and use some of them at given point of time based on inputs available.Eg. One more takes parameter a, Second model takes parameter a,b and Third model takes a,b and c and so on.When I integrate these models into App, currently getting individual MLFeatureProvider generated.Instead would like to define my own MLFeatureProvider that takes all parameters a,b,c,d etc. and invoke prediction based on number of inputs in hand.If user provided only parameter a, then would invoke model A's prediction.Presently composition seems possible solution, to have variable of MLFeatureProvider and initiate to relevant model likevar model:MLFeatureProvider = First /Second / third and then invoke corresponding prediction.Is there better way to define single MLFeatureProvider that takes say all the paramters and invoke prediction and it automatically chooses the relevant model?