Translocation error message is slightly incorrect and not helpful

[sindresorhus]

It says:

App can’t be updated if it’s running from the location it was downloaded to.

There are two other cases where it could happen:

1. The user downloads the app ZIP file, moves it to the Applications folder, unzips, and then tries launching the app.
2. The user downloads the app ZIP file to the Downloads folder, unzips, and then uses the command-line or some script to move the app to the Applications folder. (when it's moved with a script, it won't have the quarantine xattr cleared)

So the error message is technically incorrect, because with these cases, the app is not running from its download location.

I also think the error message could be more helpful. It should provide guidance what the user can do to fix the situation. The solution is simple, they must quit the app, manually move the app to the Downloads folder in Finder, and then move it back again to the Applications folder.

[zorgiepoo]

I think this is getting kind of confusing and pedantic. I don't consider these as usual cases for regular users. If the developer wants to eliminate app translocation messages, they should probably consider serving a (signed) disk image (which some developers will have their own issues against but alas). The solution that you say is "simple" is not simple to me.

There's no way for Sparkle to distinguish the common easier-to-explain case than the cases you brought up.

[sindresorhus]

Quit %1$@, move it into another folder, relaunch it from there and try again.

Does moving it to a random folder clear the quarantine xattr? I thought it had to be moved to /Applications or ~/Applications. Regardless, I think the description should make the user move the app to the /Applications folder, so that needs a two step process to cover all use-cases.

I don't consider your proposal simple/well-worded if the user has the app in a place like ~/Downloads

True, but mine does cover all the use-cases. The alternative is to detect whether it's already in the /Applications or ~/Applications and then show a different message based on that.

[zorgiepoo]

Does moving it to a random folder clear the quarantine xattr?

It is moving it to any (random) folder using Finder. If you want to be more technical, I believe Finder doesn't entirely clear the quarantine xattr but modifies it in some way to remove/flip the 'is translocated' bit.

For the immediate/soon future, I will restore the current guidance message Sparkle is intended to show (done in #2689 -- Quit %1$@, move it into your Applications folder, relaunch it from there and try again.) which already is localized. Sparkle wasn't showing this message at all and that was a bug.

The alternative is to detect whether it's already in the /Applications or ~/Applications and then show a different message based on that.

Maybe detect if the copy of the app is in /Applications and it 'looks' identical, and then guess it's running from there, but I'm not sure I'm too keen on this. It might not be 100% certainty due to app translocation preventing you from obtaining the real location of the running app. I think sandboxed apps can read arbitrary bundles in /Applications today but not sure that could ever change later on.

[sindresorhus]

It might not be 100% certainty due to app translocation preventing you from obtaining the real location of the running app.

You can detect it. Just check if it's translocated and then call NSWorkspace.shared.urlForApplication(withBundleIdentifier:). It will return the best location, which is never the translocated one.

[zorgiepoo]

And then you may have to assess it's the same version of the app and not a copy of the running app.

App translocation is not supposed to provide developers a 'supported' API to get the real/original location of that instance of the app. Sure you can ask Launch Services to return all the URLs matching a bundle ID in ranked order which I equate to guessing.

[zorgiepoo]

There is a balance here on improving the correctness of the message in the mentioned edge-case scenarios vs effort in addressing those scenarios in an odd way and asking localizers to do work vs at the extent of going against Apple wanting developers to serve a disk image. In the short term I'd like to fix the biggest issue (regression) here and see how much that helps this situation.