RE: [Gcl-devel] Re: CALL-ARGUMENTS-LIMIT and other things

[Mike Thomas]

Hi Paul.

Paul wrote:
| Mike Thomas wrote:
| > Hi Camm.
| >
| > 1. While trawling through the ANSI tester output a few days ago
| I noticed
| > that call-arguments-limit (set to 64) conflicts with the test
| in "eval.c"
| > based on the C macro MAX_ARGS (63) defined in "h/funlink.h".
| >
| > I played it safe and set call-arguments-limit to MAX_ARGS (in "eval.c")
| > which I left at 63.  This fixed the failure of SUBTRACT.3 in
| the ANSI tests:
| >
| > (LET ((ARGS NIL))
| >         (LOOP
| >           FOR
| >           I
| >           FROM
| >           1
| >           TO
| >           (MIN 256 (1- CALL-ARGUMENTS-LIMIT))
| >           DO
| >           (PUSH 1 ARGS)
| >           ALWAYS
| >           (EQL (APPLY #'- 1000 ARGS) (- 1000 I))))
| >
| > but is not necessarily a high quality design decision.  I also set
| > LAMBDA-PARAMETERS-LIMIT to MAX_ARGS.
|
|
| Looking at that, the test seems wrong -- it passes CALL-ARGUMENTS-LIMIT
| args to -, but that constant is an *exclusive* upper bound.
|
|       Paul

On that basis I'll revert that change having incorrectly assumed the limit
was inclusive.

Cheers

Mike Thomas.