Bug 3689

uim> (a)
"start a"
ERROR: bad arg to nth (see errobj)

r929 results as above. The message "ERROR: bad arg to nth" and subsequent
termination of execution is the error check what I mensioned in the commit
message of r929. Continuation without such termination as performed in r922 does
not make sense. What we should do is removing (im-lang #f) condition.

IMHO following code should generate following result.

"start a"
"FAILED"
"end e"

But revision 929 doesn't. It's not comfortable for me.

== start of a2.scm ====================================
(require "im.scm")

(define (a)
  (print "start a")
  (if (im-lang #f)
      (print "OK")
      (print "FAILED"))
  (print "end e"))

(a)

== end of a2.scm ======================================

im-lang expects an im record as argument. So (im-lang #f) is completely invalid
form. Suppose ((im *)NULL)->lang in C.

I can't imagine when (im-lang #f) will happen. Please show me actual example.

As an actual example, exec current uim-module-manager. (Note, you should make back
up of your loader.scm and installed-module.scm at first, it's not work fine at
this point.)

It works with util.scm rev922, but doesn't with 929. I don't want to debug,
because I don't 
have enough time and morale.

I decline the change since your code of r922 has a design error. See following
example.

(define-record 'foo
  '((initialized? #f)))
(define rec (foo-new))

(foo-initialized? rec)  ==> #f
(foo-initialized? #f)   ==> #f

The two results can't be distinguished by its caller. Although latter form
outputs error message for human, the caller code can't detect it.

To avoid this ambiguity, explicit #f check preceding foo-initialized? is
required as follows.

(if rec
    (foo-initialized? rec)
    (error-handlings))

I decline the design error and you should repair your own code as above. I'll
debug my codes when you found something wrong other than yours.

This problem is already fixed by the rewrite around uim-module-manager.