Hi,

At Wed, 31 Mar 2004 09:03:57 +0900,
Daniel Berger wrote in [ruby-core:02724]:
> Is this a case of "don't do that"?  Should I be casting the value passed
> to rb_time_new to ensure an integer value?  Or should
> time_new_internal() in time.c be altered?

Originally, rb_time_new() requires two time_t arguments, but is
declared in intern.h without argument declaration to just get
rid of including <time.h> in all sources.

However, now there is missing.h, so we could replace such
meaningless ANYARGS with proper prototypes.  Then, your example
will get to fail.


Index: intern.h
===================================================================
RCS file: /cvs/ruby/src/ruby/intern.h,v
retrieving revision 1.142
diff -U2 -p -d -r1.142 intern.h
--- intern.h	16 Feb 2004 06:45:31 -0000	1.142
+++ intern.h	31 Mar 2004 01:36:42 -0000
@@ -214,6 +214,6 @@ void rb_thread_interrupt _((void));
 void rb_thread_trap_eval _((VALUE, int));
 void rb_thread_signal_raise _((char*));
-int rb_thread_select(ANYARGS);
-void rb_thread_wait_for(ANYARGS);
+int rb_thread_select _((int, fd_set *, fd_set *, fd_set *, struct timeval *));
+void rb_thread_wait_for _((struct timeval time));
 VALUE rb_thread_current _((void));
 VALUE rb_thread_main _((void));
@@ -445,5 +445,5 @@ VALUE rb_struct_getmember _((VALUE, ID))
 VALUE rb_struct_iv_get _((VALUE, char*));
 /* time.c */
-VALUE rb_time_new(ANYARGS);
+VALUE rb_time_new _((time_t, time_t));
 /* variable.c */
 VALUE rb_mod_name _((VALUE));
Index: ruby.h
===================================================================
RCS file: /cvs/ruby/src/ruby/ruby.h,v
retrieving revision 1.102
diff -U2 -p -d -r1.102 ruby.h
--- ruby.h	10 Mar 2004 07:05:10 -0000	1.102
+++ ruby.h	31 Mar 2004 01:33:14 -0000
@@ -674,6 +674,6 @@ rb_special_const_p(obj)
 }
 
-#include "intern.h"
 #include "missing.h"
+#include "intern.h"
 
 #if defined(EXTLIB) && defined(USE_DLN_A_OUT)

-- 
Nobu Nakada