Interlisp
Use uname() for getparm for ARCH/MACH queries.
Previously, this was implemented in very inconsistent ways and not yet implemented for most of the platforms that we're supporting today.
The values for DOS are maintained as they were.
It's not clear to me why returning the OS name (sysname) for ARCH is a good choice. The existing code was, I suspect, attempting to duplicate the result of the SunOS/Solaris "arch" and "mach" commands and simulate it where it didn't exist. I wouldn't object to OSNAME as another parameter it returns, though, because it makes it easler for Lisp code to adapt to the underlying system.
| system | arch | mach | uname -p | uname -i | uname -m |
|---|---|---|---|---|---|
| Solaris 11.4 on a T1 | sun4 | sparc (sysinfo(SI_ARCHITECTURE)) | sparc | sun4v | sun4v |
| Solaris 11.4 (64-bit) on VirtualBox on Intel(R) Core(TM) i7-8850H | i86pc | i386 | i386 | i86pc | i86pc |
| Linux on Intel(R) Core(TM) i7-8850H | x86_64 | (none) | x86_64 | x86_64 | x86_64 |
| Linux on ARMv7 TI AM335x BeagleBone Black | armv7l | (none) | unknown | unknown | armv7l |
| macOS 10.11 on Intel core 2 duo | i386 | (none) | i386 | (none) | x86_64 |
| mac OSX on Power PC | ppc | (none) | powerpc | (none) | Power Macintosh |
| FreeBSD 12.2 on IntelPentium M processor (686-class CPU) | (none) | (none) | i386 | GENERIC | i386 |
| macOS 11 on M1 ARM | arm64 | (none) | arm | (none) | arm64 |
FreeBSD says
make(1) uses [uname -p] to set the MACHINE_ARCH variable. make(1) uses [uname -m] to set the MACHINE variable.
Solaris 11.4 says:
On sparc systems,
archreturns sun4 (historical artifact) whilearch -kreturnsuname -m. On all other systems,arch==arch -k==uname -m.
what is most useful going forward? Labels to put on benchmarks? deciding what to send to ShellCommand?
larry@Mac-mini maiko % arch
arm64
larry@Mac-mini maiko % uname -p
arm
larry@Mac-mini maiko % uname
Darwin
larry@Mac-mini maiko % uname -m
arm64
larry@Mac-mini maiko % arch -k
arch: Unknown architecture: k
How is this actually used from within Medley? The comment indicates that it isn’t used much except for whether or not ARCH is ”dos” or not and whether or not there is any result from the call ... is that still accurate?
It's not a closed universe -- we have no idea how it might have been used. We do know that it is used for checking against "dos" (lowercase).
The good thing about a 30 year gap in history is that all of the platforms that were supported, except Solaris and sort-of-DOS, are dead now. So any legacy code checking this for things like MIPS or 68k or Irix or AIX ... wouldn't care what the new values are. So I think it is best if we can get it right now. It is also worth noting that many of our current platforms are mis-reported by this, like any ARM / Aarch64 platforms.
getting something that is meaningful for labelling benchmarks would be good.
No, it doesn't tell you which OS you're running -- could be x86_64 darwin, or x86_64 linux, and they need different copy/paste commands. It also tells you "X", but they're all "X', so that's no help. I think we can probably settle on something for "MACH" and "ARCH", but Ron needs "OSNAME" -- and we'll get that from the OS we compiled on. I've got a lot of things I'm working on.
Ron should be using new subrs that let Maiko correctly implement copy/paste. Doing otherwise simply won't work on all platforms going forward as shell commands aren't always available to do this.