Problem with equal-hash? -- Segmentation fault: 11

[Takashi Kato]

Reported on Github.

https://github.com/ktakashi/sagittarius-scheme/issues/1

[Takashi Kato]

eqv? of record causes stack overflow. The following is a simple example:

(import (rnrs))
(define-record-type (<pare> kons pare?)
  (fields (mutable a kar set-kar!)
      (mutable d kdr set-kdr!)))

(let ((a (kons 'a 'b))
      (b (kons 'a 'b)))
  (set-kdr! a a)
  (set-kdr! b b)
  (eqv? a b))
;; -> SEGV

[Takashi Kato]

Both R6RS and R7RS say that comparison of record by eqv? returns #t only if the memory location is the same. So basically, the error case should always return #f.

NB: this also means equal? must also return #f for the sake of R6RS. (or more precisely, the R6RS test suites.)

[Takashi Kato]

eqv? shouldn't inspect record fields. (Fixes #182) Comforming R6RS equal? strictly (breaking backward compatibility)

→ <<cset a40b2b1018da>>