I found the path to "Choose Template" which was not obvious either. Something more like "Select Project Type" would have been more definitive to me. There are lots of reasons to allow AIs to be a part of the development process. And I appreciate and use them. But technically, I prefer to create the project myself and select the type of project I want and then even create some domain vocabulary in parts of the starting project to help manifest suggestions to the AI that allow it to readily infer the domain and the verbs I want to design around.

I have both com.apple.security.device.audio-input and com.apple.security.device.microphone present in entitlements already, and I have NSMicrophoneUsageDescription in info.plist as well. My app does not appear in "Privacy & Security -> Microphone" settings to grant permission for microphone use though. And again, this is "Mac (Designed for iPad)" build which is only going to work on Apple Silicon Macs. My laptop is such a device. I can run the app out of Xcode. The Sandbox settings do not appear when I select "Mac (Designed for iPad)" instead of "Mac Catalyst" or just "Mac" builds.

Ultimately, there seems to be quite a bit of integration into "music" playing, but I need my audio to come through AVAudioSession/AVAudioEngine because it's raw, streaming audio. I have meta data to display and so the now-playing Deck seems like the best place to put that meta data because there is already so much provided with the play/pause buttons and the album cover image. If I use .playAndRecord, the now-playing dev navigation button does not appear. If I use .playback, the button does not appear unless I create an error by also trying to use .defaultToSpeaker. What is it that I cannot get the now-playing deck navigation button to appear?

I am not sure if there is just no-one doing much with car play audio, or if there is something unclear with the above text. What I have found is that I cannot get the navigation to the now-playing deck to appear on the radio without using .playback category and also specifying .defaultToSpeaker which is not an acceptable combination. I have to use session.setActive(false) before trying to set the category, and then catch the error thrown, and then use .setActive(true) after the error and then things work. The is the only way I seem to be able to set the artist, album title and artwork on the radio to make it visible to the users. do { try session.setActive(false) try session.setCategory(.playback, mode: .default, options: [ .defaultToSpeaker, .allowBluetoothA2DP, .mixWithOthers ]) try session.setActive(true) logRoute("setCategory(.playback) [CarPlay]") } catch { try session.setActive(true) logRoute("setCategory(.playback) [CarPlay] failed: \(error)") } This block just doesn't seem like the right thing to do. .playAndRecord does not seem to allow anything about the now-playing deck to be configured and the navigation button does not appear. I have to push the now-playing view to see it, but it's blank with none of the details I've set with code like: let info = nowPlayingInfo() MPNowPlayingInfoCenter.default().playbackState = dataValue != nil ? .playing : .stopped MPNowPlayingInfoCenter.default().nowPlayingInfo = info

Why is there still no llvm support for apple silicon? It's been over 4 years and swift has had hundreds of changes (the compiler still has basic issues with not being able to complete compilation, so it still needs work), but we have to agree, at some level, that it would be much better for LLVM to support apple hardware, and for apple's tooling to actually build with LLVM as the back end such that all kinds of compatibility for exploring best of class language use on apple silicon would be possible.

Ultimately this problem just points out how detrimental to progress and usability it was to omit the need of some statement terminator, such as a semi-colon. Now we have a completely broken and incapable recursive decent parser that just cannot arrive at a usable conclusion on the appropriate problem line of code. So we can to waste 100s of hours of development time, each, trying to work around this nonsense of a development environment. It is literally 3 years later/after this ticket was first posted and this is still happening.

Lazy initialization implies non-concurrent access which will be another thorn in our side as Swift matures. There must be some form of planned, ordered execution that occurs in a predictable way so that we don't have to sprinkle lazy initialization around everywhere. One of the primary problems I am confronting is in the use of SwiftData and making sure that there is at least one tuple in place for particular models that store and convey settings across the application. The use of .modelContext() creates a lot of lazy initialization dependency problems because it doesn't really allow for the above initialization of at least one tuple to readily happen because of this: Main actor-isolated property 'mainContext' can not be referenced from a non-isolated context

There seems to still be issues with how selection works, but in your case, once you designate a tag. the selection type needs to be Int to get that tag value, or at least that's what seemed to work for me based on a tutorial I've lost the link to.

The swift compiler, as a recursive descent parser is terribly slow at resolving broken code. It really needs to be replaced with an LALR1 compiler using lex/yacc with the swift syntax more fully described in the grammar instead of their being so much context sensitive choices that require the compiler to unwind and recompile from so many different levels to try and discern what is meant by the content. Modern day compilers should be complaining about syntax errors much more quickly. Certainly named parameters make it a lot harder to "figure out" the polymorphic resolve of the exact method. But realistically I am having to wait 5min or more for some problems that are only reported as "The compiler is unable to type-check this expression in reasonable time; try breaking up the expression into distinct sub-expressions" The SwiftUI chaining of .onChange(), and other things with code in them create the biggest barrier here. Something different needs to happen here that moves the base syntax resolution to happen first and then have polymorphic binding result methods. That may already be what's being done, I don't know, but I just am spending day after day commenting out lines of code trying to find the syntax error and get a real error message. It's keeping me from being productive and pushing me to just stop and give up again on these apps I keep trying to put together but have to rewrite because each time I try again, I have a complete new set of APIs, language, libraries etc. to deal with.

I am working on a new app in iOS 17.2 on my iPhone. The location privacy settings for "always", "when running" and "description" all all in place. However, the settings for location access permissions only reveals "Never" and "When I share" as options. This seems to be where there is a problem. Another App I worked on before iOS17, but did change to use the beta, shows all 4 selections and I can move it to "always", and then my location appears in that app.

The ++ and -- operator removal is about as childish of a choice as one might find in language design. There are countless examples of how these operators are priceless. let elems = pkt.split(separator: ","); //$GPRMC,024613,V,3609.9338,N,09556.3937,W,000.0,000.0,160204,004.6,E*7E if( gpsDataType == "GPRMC," ) { let time = elems[0] let warn_V_A = elems[1] let lat = elems[2] let latdir = elems[3] let lon = elems[4] let londir = elems[5] let speed = elems[6] let course = elems[7] let date = elems[8] let magvar = elems[9] let dirs = elems[10].split(separator: "*") let vardir = dirs[0] let check = dirs[1] let latitude = parseLat( String(lat), dir: String(latdir) ) let longitude = parseLon( String(lon), dir: String(londir)) self.latitude = latitude self.longitude = longitude self.heading = Int(Double(course)!) self.speed = Int(Double(speed)!) } Why do I have code all these stupid indexes with constants? Why can I not use [idx++] on every index so that I don't have to worry about what line it is, and then for other GPS data, like $GPGGA, I can just copy the lines above that I need to use for that statement and not care at all what the indexes are.

Isn't the CGRect passed into draw the cropping rect for where drawing needs to happen? It seems like you should be drawing into an offline buffer and then just blitting from that drawing into the CGRect area.

This appears to be an order of execution problem. The issue is that there are two many pointers to too many things. the delegate pattern is creating lots of issues because I need to create a delegate object and pass the view object to it. Or, I have to make all the objects be delegates themselves. This greatly affects the flexibility of the software system design.

You need 'public' on those variable declarations