site_archiver@lists.apple.com Delivered-To: darwin-dev@lists.apple.com On Mar 12, 2009, at 6:10 AM, Terry Lambert wrote: For a drag and drop type install, LIPO isn't an option. - Jim _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-dev mailing list (Darwin-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-dev/site_archiver%40lists.appl... The place where this would be important today is on a machine capable of running 64 bit Intel binaries, but for which only the system framework is available, and not other frameworks. Since we changed the 64 bit binary cpu type number by adding a "has 64 bit frameworks bit" to it, when we added framework support, if you link a new binary, the 386 version of the binary is going to be preferred by the older version of the OS anyway, so you'd be solving a problem that doesn't exist in the first place. The other place it could be important is if you ported over your 32 bit KEXT or daemon that talked to it, and decide, for whatever reason that you wanted to still communicate ponter and long values between it and user space instead of using sized types. The other place where I wanted to control this is where my GUI app, which runs as a 32-bit process on Leopard (which can be influenced by LSMinimumSystemVersionByArchitecture), launches a helper (today via exec), and I'd like to run the helper as 32-bit on Leopard too, mainly because I don't want to be the reason the 64-bit frameworks are loaded in. But you could also LIPO at install time with your install tool or the binary itself when it saw the results of the uname(2) system call, or you could also write a small script to wrap the use of the arch(1) command, or you could use the posix_spawn(2) extensions documented in posix_spawnattr_setbinpref_np(2) to reexec yourself, if you didn't mind running the 64 bit binary, saying "whoops! I should be 32 bit!" and restarting yourself with a different preference. It looks like using posix_spawn solves the problem. (Previously I was using fork/exec because the code was originally written to target 10.4.x.) This email sent to site_archiver@lists.apple.com