Re: [MIT-Scheme-devel] x86-64 assembler problem

[Taylor R Campbell]

   Date: Wed, 27 Apr 2011 15:15:22 -0700
   From: address@hidden (Matt Birkholz)

   > From: Taylor R Campbell <address@hidden>
   > Date: Wed, 27 Apr 2011 20:01:33 +0000
   > 
   > [...] you can probably work around it by breaking your code into
   > smaller parts.

   I'll try that.

   > How did you trigger this?  Just by building the current state of
   > <git://birkholz.chandler.az.us/~matt/mit-scheme.git>?

   No, you will want to remove the "#| Kludge:" comment to break it.

Yes, try splitting that definition up like this:

(define gtk-key-translation-list '())

(define (define-gtk-key symbol enumerand)
  (set! gtk-key-translation-list
        (cons (cons enumerand symbol) gtk-key-translation-list))
  unspecific)

(define-gtk-key 'BRACELEFT (C-enum "GDK_braceleft"))
(define-gtk-key 'BAR (C-enum "GDK_bar"))
;; &c.

(define gtk-key-translation-table
  (sort! (list->vector gtk-key-translation-list)
         (lambda (a b) (< (car a) (car b)))))

As is, CONS and presumably C-ENUM will be open-coded and involve no
procedure calls, so there will be a giant contiguous chunk of code,
which LIAR won't handle well.  (This will also exhibit another danger
of LIAR -- it only heuristically checks for heap overflows, so if you
use enough open-coded allocation primitives, you will lose.)