RE: [ANNOUNCE] Lua 5.0 (pre-release) now available

[Benoit Germain <bgermain@...>]

Ok, I've progressed a bit. Oh, I am working with the final release of
course.

So, I just changed INT_MAX to 32767 in the file (BTW, the next prc-tools
release will fix this bug).

I also changed the core's code a bit to accommodate for various
setjmp/longjmp implementations, because the palm OS doesn't have these exact
functions. It defines instead ErrJumpBuf, ErrJongJump and ErrSetJump. So I
have created a file etc/luser_exception.h in which one can redefine the
setjmp/longjmp mechanism, this file being included in ldo.c where the macro
definitions default to the standard setjmp/longjmp mechanism. Tell me if you
are interested.

Now I am back to a single remaining sizeof(int) problem, which comes from
lua code:

lobject.c: In function `luaO_log2':
lobject.c:63: warning: right shift count >= width of type
lobject.c:64: warning: right shift count >= width of type

I just changed the code by adding 2 lines as follows:

int luaO_log2 (unsigned int x) {
  static const lu_byte log_8[255] = {
    0,
    1,1,
    2,2,2,2,
    3,3,3,3,3,3,3,3,
    4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,
    5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,
    6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
    6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,
    7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
    7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
    7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
    7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7
  };
  #if INT_MAX > 32767 /* ################# */
  if (x >= 0x00010000) {
    if (x >= 0x01000000) return log_8[((x>>24) & 0xff) - 1]+24;
    else return log_8[((x>>16) & 0xff) - 1]+16;
  }
  else
  #endif /* ################## */
  {
    if (x >= 0x00000100) return log_8[((x>>8) & 0xff) - 1]+8;
    else if (x) return log_8[(x & 0xff) - 1];
    return -1;  /* special `log' for 0 */
  }
}

Then, in ltable.c:

ltable.c: In function `luaH_mainposition':
ltable.c:108: warning: cast from pointer to integer of different size
ltable.c:110: warning: cast from pointer to integer of different size

This is because sizeof(int)==2, whereas sizeof(void*)==4
I suppose that this is not important because even if data is lost when
hashing pointers to ints, the tables will still work fine, but maybe there
is a possibility to silence that warning ?

After that, the lua core compiles fine except for strtod, strcspn, strcoll
and which don't exist in the palm os headers. I'll solve that later at link
time with a host program.

Onward to the standard libraries then:
lauxlib and lbaselig being quite necessary, I had to adapt them to the
FILE-less system, loadfile-related stuff does not compile -> I added a
LUA_HAS_FILE directive that neutralizes dofile() and loadfile().
Also, no console, so no stderr, no fprintf -> I changed the default behavior
of callalert() to be defined by a user macro LUA_DEFAULT_ALERT(string)
defaulting to the current implementation.
For the rest, obviously I can't use liolib.c, so I only have to avoid
compiling it. The other files are ok.



Thanks to everybody for your rapt attention :-)


Benoit.


-----Message d'origine-----
De : Benoit Germain [mailto:bgermain@ubisoft.fr]
Envoyé : lundi 14 avril 2003 16:04
À : Multiple recipients of list
Objet : RE: [ANNOUNCE] Lua 5.0 (pre-release) now available 


the INT_MAX I get is 2147483647, defined in
/lib/gcc-lib/m68k-palmos/2.95.3-kgpd/include/limits.h, so there might be a
problem with those files. But it is surprising, because I am not the first
to use this development suite.


-----Message d'origine-----
De : Roberto Ierusalimschy [mailto:roberto@inf.puc-rio.br]
Envoyé : lundi 14 avril 2003 14:56
À : Multiple recipients of list
Objet : Re: [ANNOUNCE] Lua 5.0 (pre-release) now available 


> so obviously there is a problem, but I can't investigate right now.

Can you send me the value of the constant INT_MAX in your program?
(defined in limits.h)

Thanks,

-- Roberto