Fix CLOS, MAKE-LOAD-FORM, and TYPE-OF ANSI test failures

[blakemcbride]

Fix CLOS, MAKE-LOAD-FORM, and TYPE-OF ANSI test failures

Summary

Resolves 32 ANSI test failures across eight areas of ABCL's CLOS,
type system, and file-compiler behavior. All originally-targeted
tests pass, with no regressions against the baseline test suite.

Builds on (and cherry-picks) the MAKE-LOAD-FORM data-flow ordering
commit, and adds seven further fixes.

Area	Tests fixed
MAKE-LOAD-FORM.ORDER.3–14	12
DEFINE-METHOD-COMBINATION-LONG.04.2, .11.2, .11.3	3
DEFGENERIC.ERROR.1/2/3	3
CALL-NEXT-METHOD.ERROR.1/2	2
DEFSTRUCT.ERROR.3/4	2
FIND-CLASS.24	1
MAKE-CONDITION.3/4	2
TYPE-OF.1/4 (and TYPE-OF.1-relaxed)	3
DEFCLASS.FORWARD-REF.1/2/3 (kept green)	3

Changes

1. MAKE-LOAD-FORM data-flow ordering (ORDER.3–14)

Cherry-picked from the make-load-form branch. The file compiler now
emits creation and initialization forms through a per-fasl instance
vector sys::*fasl-instances*, with a two-phase dependency walker that
honors CLHS's "creation before any reference" and "init ASAP after
creation" rules. *FASL-VERSION* is bumped 43 → 44. See the
make-load-form PR description for full detail on this portion.

2. DEFINE-METHOD-COMBINATION long form (clos.lisp)

Three independent bugs in the long-form define-method-combination
expansion:

• :order / :required evaluated as forms, not quoted symbols.
  canonicalize-method-group-spec was emitting :order ',order and
  :required ',required-p, so a group spec like
  ((method-list * :order order)) substituted the literal symbol
  order rather than the value of the lexical variable. CLHS
  specifies these are forms to be evaluated at call site.
• Optional parameters "in excess" of the generic function's
  optional count were bound to NIL instead of their initforms.
  Rewrote the optional-parameter handling in
  method-combination-type-lambda-with-args-emf to emit an
  initform binding for excess optionals, and to bind excess
  supplied-p parameters to NIL.
• Keyword supplied-p parameters leaked the member tail. The
  same binding symbol was used for both the boolean supplied-p and the
  (cadr …) accessor, so user code saw e.g. (:Z1 4) instead of T.
  Split these via a private gensym for the member result;
  supplied-p now binds to (not (null …)).

3. CALL-NEXT-METHOD applicable-methods check (clos.lisp)

AMOP requires that when CALL-NEXT-METHOD is passed an argument list,
those arguments must produce the same set of applicable methods as the
original call. ABCL previously silently re-dispatched on whatever
arguments it received.

Added *call-next-method-gf* / *call-next-method-applicable-methods*
specials bound by the slow-method-lookup path, plus a
cnm-check-applicable-methods helper invoked from the
flet-call-next-method expansion when explicit CNM args are given. The
emfun is wrapped so the binding survives effective-method caching.

4. DEFCLASS top-level compile-time recognition (clos.lisp)

FIND-CLASS.24 requires that a (defclass test-class …) appearing at
top level in a compiled file be visible to macros invoked later in the
same compilation (via (find-class 'test-class nil env)). The defclass
macro now:

• at :compile-toplevel, registers a placeholder
  forward-referenced-class so later compile-time find-class calls
  succeed;
• at :load-toplevel, clears that placeholder only when it has no
  direct subclasses (i.e. it really is our own placeholder, not a
  legitimate forward reference from a previously-defined subclass),
  so ensure-class then goes through the primitive path and avoids
  a MOP superclass-validity assertion that built-in classes like
  STREAM fail;
• at :execute, does not touch existing forward-references
  (fixes the DEFCLASS.FORWARD-REF.* regression introduced by an
  earlier version of this fix).

5. TYPE-OF vs. TYPEP consistency (NilVector + SGF)

TYPE-OF.1 / .4 require that for every object x, (type-of x) be
a subtype of every built-in type that x is TYPEP. Two specific
instances failed this invariant in ABCL:

• NIL-vector ↔ BASE-STRING. A (simple-array nil (*)) reported
  (typep x 'base-string) → T, but
  (subtypep '(nil-vector n) 'base-string) → NIL, T because the
  upgraded element types (nil vs character) are disjoint per CLHS
  4.2.3. NilVector.typep now returns NIL for BASE-STRING /
  SIMPLE-BASE-STRING. (STRING remains T — nil is a subtype of
  character, so a NIL-vector is legitimately a string.)
• STANDARD-GENERIC-FUNCTION ↔ COMPILED-FUNCTION.
  FuncallableStandardObject.typep previously returned T for
  COMPILED-FUNCTION when the installed dispatcher happened to be
  compiled; this is an implementation detail of the funcallable
  instance, not a property of the object itself. Returning NIL
  matches SBCL/CCL behavior. Because jvm-compile used the old T
  answer as an early-return signal, it now explicitly skips
  funcallable-standard-object definitions (they have no LAMBDA
  expression to work from).
• subtypep bugfix in the string-array branch of
  csubtypep-array: when the second type was a bare symbol like
  nil-vector (no dimension spec), the size check compared
  (car i1) against nil instead of treating the missing dimension
  as *. (subtypep '(nil-vector 0) 'nil-vector) now returns
  T, T.

Files changed

Beyond the cherry-picked make-load-form commit:

• src/org/armedbear/lisp/clos.lisp
  • defclass macro: compile-time placeholder + guarded
    load-time clear.
  • canonicalize-method-group-spec: evaluate :order /
    :required at call site.
  • method-combination-type-lambda-with-args-emf: excess-optional
    initform binding; split key supplied-p from member tail.
  • *call-next-method-gf*, *call-next-method-applicable-methods*,
    cnm-check-applicable-methods; slow-method-lookup emfun wrap.
• src/org/armedbear/lisp/defstruct.lisp
  • (DEFSTRUCT.ERROR.3/4 corrections.)
• src/org/armedbear/lisp/compiler-pass2.lisp
  • jvm-compile: short-circuit on funcallable-standard-object.
• src/org/armedbear/lisp/FuncallableStandardObject.java
  • typep: return NIL for COMPILED-FUNCTION.
• src/org/armedbear/lisp/NilVector.java
  • typep: return NIL for BASE-STRING / SIMPLE-BASE-STRING.
• src/org/armedbear/lisp/subtypep.lisp
  • csubtypep-array string-branch: treat missing dimensions as *.

Test plan

• All 28 originally-targeted tests pass
  (MAKE-LOAD-FORM.ORDER.3–14, DEFGENERIC.ERROR.1/2/3,
  CALL-NEXT-METHOD.ERROR.1/2,
  DEFINE-METHOD-COMBINATION-LONG.04.2/.11.2/.11.3,
  DEFSTRUCT.ERROR.3/4, FIND-CLASS.24, MAKE-CONDITION.3/4,
  TYPE-OF.1/4).
• DEFCLASS.FORWARD-REF.1/2/3 continue to pass (regression guard
  on the DEFCLASS compile-time change).
• All 33 DEFINE-METHOD-COMBINATION-LONG.* tests pass.
• All 14 MAKE-LOAD-FORM.ORDER.* tests pass.
• Full cl-test suite: 21932 tests, failure count drops from
  119 → 113. No new CLOS-related failures; remaining failures
  are pre-existing and unrelated (pathname/file, FORMAT,
  LOOP.1.40–43, etc.).

Compatibility

Fasl layout changed (MAKE-LOAD-FORM portion) — *FASL-VERSION* was
already bumped 43 → 44, and older .abcl fasls must be recompiled.
No other public API changes; internal additions
(*call-next-method-gf*, cnm-check-applicable-methods,
%fasl-* helpers) are in SYSTEM / EXT.

[blakemcbride]

Duplicated in pull request 743