On 25.10.2005., at 21:02, Alvaro Lopez Ortega wrote:
Marko Mikulicic wrote:
> in config.h I see
> /* Number of bits in a file offset, on hosts where this is
settable. */
> /* #undef _FILE_OFFSET_BITS */
> someone knowns well the autoconf/automake
I think I do, but it doesn't make sense to me..
configure.in uses AC_SYS_LARGEFILE, which is a function that defines
_FILE_OFFSET_BITS and add a compilation flag in order to activate
the large file support.
Could you please check what is detecting AC_SYS_LARGEFILE?
I think the problem is in the configure stuff, the code is, AFAIK,
alright.
I'm afraid I have to say that there is some subtle interpretation
error on what AC_SYS_LARGEFILE seems to mean.
it generates the following code in configure:
----------------------------------------------
echo "$as_me:$LINENO: checking for _FILE_OFFSET_BITS value needed
for large files" >&5
echo $ECHO_N "checking for _FILE_OFFSET_BITS value needed for large
files... $ECHO_C" >&6
if test "${ac_cv_sys_file_offset_bits+set}" = set; then
echo $ECHO_N "(cached) $ECHO_C" >&6
else
while :; do
ac_cv_sys_file_offset_bits=no
cat >conftest.$ac_ext <<_ACEOF
/* confdefs.h. */
_ACEOF
cat confdefs.h >>conftest.$ac_ext
cat >>conftest.$ac_ext <<_ACEOF
/* end confdefs.h. */
#include <sys/types.h>
/* Check that off_t can represent 2**63 - 1 correctly.
We can't simply define LARGE_OFF_T to be 9223372036854775807,
since some C++ compilers masquerading as C compilers
incorrectly reject 9223372036854775807. */
#define LARGE_OFF_T (((off_t) 1 << 62) - 1 + ((off_t) 1 << 62))
int off_t_is_large[(LARGE_OFF_T % 2147483629 == 721
&& LARGE_OFF_T % 2147483647 == 1)
? 1 : -1];
int
main ()
{
;
return 0;
}
_ACEOF
----------------------------
the phrase "checking for _FILE_OFFSET_BITS value needed for large
files" means to me "ensure that we have to define _FILE_OFFSET_BITS
explicitly so that off_t would be defined to be 64-bit wide". A proof
is in the checking code itself. Let's see it by itself (the main is
empty):
---------------------------------------
#include <sys/types.h>
/* Check that off_t can represent 2**63 - 1 correctly.
We can't simply define LARGE_OFF_T to be 9223372036854775807,
since some C++ compilers masquerading as C compilers
incorrectly reject 9223372036854775807. */
#define LARGE_OFF_T (((off_t) 1 << 62) - 1 + ((off_t) 1 << 62))
int off_t_is_large[(LARGE_OFF_T % 2147483629 == 721
&& LARGE_OFF_T % 2147483647 == 1)
? 1 : -1];
--------------------------------------
so, LARGE_OFF_T is an expression that yields to (off_t)
9223372036854775807
which is (off_t)0x7fffffffffffffff, 63 least significant bits set to
one and the most significant bit set to zero.
the macro is a work around for some compilers. we can write it like
this:
int off_t_is_large[((off_t)0x7fffffffffffffff % 0x7fffffed == 0x2d1
&& (off_t)0x7fffffffffffffff % 0x7fffffff == 1)
? 1 : -1];
if off_t is 32 bit value then the cast evaluates only to the leas
significant bits:
(gdb) print (unsigned long)0x7fffffffffffffff
$4 = 0xffffffff
so the remainder of the expressions are different:
as you can see in gdb:
(gdb) print (unsigned long long)0x7fffffffffffffff % 0x7fffffed
$6 = 0x2d1
(gdb) print (unsigned long)0x7fffffffffffffff % 0x7fffffed
$7 = 0x25
the rest of the trick is to generate a declaration of an array with
negative dimesions in order to trigger a compilation error
in case the value is truncated by the cast, yeilding to:
large.c:9: error: size of array 'off_t_is_large' is negative
since darwin defaults to 64 bit off_t this succedes without the
special macro.
(I ignore why the simply didn't do (off_t)0x7fffffffffffffff >> 62 ==
1 ? 1 : -1 ... but probably there are some strange compilers out
there.)
Historically programs were built assuming a 32 bit off_t. In order to
preserve backwards compatibility one should explicitly
define _FILE_OFFSET_BITS or LARGE_FILE (in some platforms like AIX)
so that unistd.h can define off_t to have 64 bits and eventually define
macros for using correct syscalls etc. On darwin there was no need
for backward incompatiblity so it is natively implemented as:
./sys/_types.h: typedef __int64_t __darwin_off_t; /*
[???] Used for file sizes */
So this is my motivation to think that you were using incorrectly the
presence of _FILE_OFFSET_BITS,
also because on AIX for example it would define another macro and not
_FILE_OFFSET_BITS.
is there a list of platforms where cherokee was tested on?
-------------------------
Now we can talk about the solution. Anyone know how to get autoconf
define a macro which is true even on platforms
where large file is default?
-- marko
PS: wouldn't be better to use reply-to to the list instead of having
to use "replay-all" in the mail client?
_______________________________________________
Cherokee mailing list
[email protected]
http://www.alobbs.com/cgi-bin/mailman/listinfo/cherokee
