PRINT-LISP-INFORMATION needs updating

[masinter]

it says Medley 2.0, always i386, memory size 0, make-init date, and patch files?

It's partly from PRINT-LISP-INFORMATION but other sources should be included?

[nbriggs]

If the Lisp code does (UNIX-GETPARM "MACH") then it will get "sparc" (on any SPARC system), "386" on any DOS system, or "i386" on any macOS system, otherwise "mc68020". This is pretty much useless.

CMLENVIRONMENT's MACHINE-TYPE is the only code that uses that call. It's supposed to return "the generic name of the computer hardware on which Common Lisp is running". On D-Machines it would have returned "Xerox 1108", "Xerox 1186", or "Xerox 1132". The examples in the hyperspec are "DEC PDP-10", or "DEC VAX-11/780", or "Symbolics LM-2" -- so the "Xerox 11xx" names are in line with that, but I don't know what to do about detecting everything else. The exact cpu model is not generic enough, but might do for the MACHINE-VERSION call. Perhaps "Intel x86" or "Intel x86_64", or whatever you get on an AMD, ARM, RISC-V is the right thing -- sort of what we currently have in the MAIKO_ARCH_NAME which we get through (UNIX-GETPARM "ARCH") which returns one of the following:

- #define MAIKO_ARCH_NAME "WebAssembly"
- #define MAIKO_ARCH_NAME "x86_64"
- #define MAIKO_ARCH_NAME "arm"
- #define MAIKO_ARCH_NAME "arm64"
- #define MAIKO_ARCH_NAME "x86"
- #define MAIKO_ARCH_NAME "PowerPC"
- #define MAIKO_ARCH_NAME "RISC-V"
- #define MAIKO_ARCH_NAME "SPARC"
- #define MAIKO_ARCH_NAME "Motorola68K"

This doesn't line up with the only use of that call, which is in LLKEY where it does (STREQUAL "dos" (UNIX-GETPARM "ARCH")). We do have (UNIX-GETPARM "OSNAME"), which doesn't return "dos", but can return "Windows" and many other OS names.

I think we need these functions (from the CLHS):

• (lisp-implementation-type) => "Medley Lisp"
• (lisp-implementation-version) => ? (git version? "3.5.1" ? makesys date? Maiko version? something else?)
• (machine-type) => see above
• (machine-version) => see above
• (machine-instance) => could be hostid, serial number, UUID, or just nil
• (software-type) => supposed to be the supporting software name -- OS name (Linux, macOS, Haiku, FreeBSD) and perhaps Maiko shows up here, say, "osname/Maiko"
• (software-version) => OS version (e.g., "Ubuntu 20.04.6 LTS", "Sonoma 14.8.7", ?, "14.4-RELEASE-p4", and a Maiko version here if we put it in the software-type)
• (short-site-name), (long-site-name) => who knows... nil? user settable? it is supposed to be the physical location of the system, which doesn't seem too useful.

If we can figure out how to retrieve the information we think should be returned, for each OS we support, I don't think there's anything holding us back from modifying the UNIX-GETPARM keys and results.

[pamoroso]

On GNU CLISP short-site-name and long-site-name return uname -n and uname -a or the value of the environment variable ORGANIZATION if bound. On my Linux box:

[1]> (short-site-name)
"lispmachine"
[2]> (long-site-name)
"Linux lispmachine 6.8.0-117-generic #117-Ubuntu SMP PREEMPT_DYNAMIC Tue May  5 19:26:24 UTC 2026 x86_64 x86_64 x86_64 GNU/Linux"

(let ((cache nil))
(defun short-site-name ()
  (if cache cache
      (setq cache
            (or (getenv "ORGANIZATION")
                (with-open-stream (s (make-pipe-input-stream "uname -n"))
                  (read-line s)))))))

(let ((cache nil))
(defun long-site-name ()
  (if cache cache
      (setq cache
            (or (getenv "ORGANIZATION")
                (with-open-stream (s (make-pipe-input-stream "uname -a"))
                  (read-line s)))))))

[masinter]

What would you would want PRINT-LISP-INFORMATION to output when reporting the timings of the Gabriel benchmarks?

I didnt make as much headway on this than I expected. I'm leaning at present to setting up environment variables to pass on as the results in the cases where there arent clear answers.
This will at least give us a way of giving answers that aren't just nonsense.

• (lisp-implementation-type) => "Medley Lisp"

Using "Medley Lisp" has the disadvantage of it not really being a query term. So I'd vote for
"Medley Interlisp"

• lisp-implementation-version) => ? (git version? "3.5.1" ? makesys date? Maiko version? something else?)

There are released versions and development versions. Released versions just need a date.
Development versions could be 'who made it' and date/time.

• (machine-type) => see above
• (machine-version) => see above

I think the discussion (above)[https://github.com//issues/2618#issuecomment-4483106196] leads to using MAIKO_ARCH_NAME = (UNIX-GETPARM "ARCH") for machine-type.
I'm not sure more details are readily available.

• (machine-instance) => could be hostid, serial number, UUID, or just nil

hostname? user name? If you wanted to tell something which machine you used to run a benchmark, I think informally I might say Larry Masinter's Macbook Pro aquired 2025?

(software-type) => supposed to be the supporting software name -- OS name (Linux, macOS, Haiku, FreeBSD) and perhaps Maiko shows up here, say, "osname/Maiko"

The distinction between hardware and software is a little fuzzy? Anyway, machine-type and software-type overlap. Maybe we could set up some environment variables and just pass this through.

• (software-version) => OS version (e.g., "Ubuntu 20.04.6 LTS", "Sonoma 14.8.7", ?, "14.4-RELEASE-p4", and a Maiko version here if we put it in the software-type)

same as software-version.

• (short-site-name), (long-site-name) => who knows... nil? user settable? it is supposed to be the physical location of the system, which doesn't seem too useful.

SBCL on my mmac returns NIL.