[clisp-list] Re: Trying to build with syscalls support

[Thomas F. B. <tfb@OCF.Berkeley.EDU>]

Sam Steingold writes:
 > > * Thomas F. Burdick <gso@BPS.Orexryrl.RQH> [2003-10-27 17:19:25 -0800]:
 > >
 > > I'm trying to compile a clisp with the syscalls module, so I can use
 > > posix:stream-lock instead of the gross hack I currently use, but
 > > compiling 2.31 on Linux/x86 dies.  I must be doing something wrong,
 > > because I'd imagine this would be the most tested platform.
 > 
 > indeed...
 > 
 > >   calls.c: In function `C_subr_posix_file_stat':
 > >   calls.c:553: parse error before `file'
 > >   calls.c:556: `file' undeclared (first use in this function)
 > >   calls.c:556: (Each undeclared identifier is reported only once
 > >   calls.c:556: for each function it appears in.)
 > >   calls.c:557: warning: left-hand operand of comma expression has no effect
 > >   calls.c:561: `buf' undeclared (first use in this function)
 > >   calls.c:578: warning: left-hand operand of comma expression has no effect
 > >   make[1]: *** [calls.o] Error 1
 > >   make[1]: Leaving directory `/tmp/build/clisp-2.31/build-myclient/syscalls'
 > >   make: *** [syscalls] Error 2
 > 
 > I bet your GCC is not 3 (i.e., not C99).

Nope, it's 2.95

 > Sorry.
 > Please try the appended patch.

Thanks, that almost does it.  After applying that patch, the compiler
complained that clisp.h was declaring a global variable after a
function definition, when it was included from calls.c.  With the
appended patch, it compiles all the way, and the produced clisp seems
to kind-of work.

Actually, it dies in "make testsuite" with a segfault on this test:
(LET ((H (MAKE-HASH-TABLE :TEST `(,(LAMBDA (A B) (LIST (LIST '= A B)) (= A B)) . ,(LAMBDA (X) (LET ((Z (SXHASH (COERCE X 'DOUBLE-FLOAT)))) (LIST `((HASH ,X) => ,Z)) Z)))))) (LOOP :FOR I :FROM 0 :TO 1000 :DO (SETF (GETHASH I H) (FORMAT NIL "~r" I))) (LOOP :FOR I :FROM 0 :TO 1000 :UNLESS (STRING= (GETHASH (FLOAT I 1.0d0) H) (GETHASH (FLOAT I 1.0s0) H)) :COLLECT I)) make[1]: *** [tests] Segmentation fault

The non-syscalls build fails here, cleanly:
(IF *NO-ICONV-P* "AZ" (CONVERT-STRING-FROM-BYTES '#(255 254 65 0 13) (MAKE-ENCODING :CHARSET "utf-16" :INPUT-ERROR-ACTION #\Z))) 
ERROR!! "
*** - Character #\uFEFF cannot be represented in the character set CHARSET:ISO-8859-1

I'm a little uncomfortable with the resulting build, but I'll try
testing the actual application with it.

--- syscalls/calls-orig.c       Tue Oct 28 09:42:18 2003
+++ syscalls/calls.c    Tue Oct 28 09:42:46 2003
@@ -5,6 +5,7 @@
  */
 
 #include "config.h"
+#include "clisp.h"
 
 #if defined(TIME_WITH_SYS_TIME)
 # include <sys/time.h>
@@ -39,8 +40,6 @@
 # define UNIX_CYGWIN32
 # undef UNICODE
 #endif
-
-#include "clisp.h"
 
 /* #define DEBUG */
 #if defined(DEBUG)

-- 
           /|_     .-----------------------.                        
         ,'  .\  / | No to Imperialist war |                        
     ,--'    _,'   | Wage class war!       |                        
    /       /      `-----------------------'                        
   (   -.  |                               
   |     ) |                               
  (`-.  '--.)                              
   `. )----'