RFH: compile-time assertions now broken in GCC

[Thorsten Glaser <tg@debian.org>]

Hi,

https://bugs.launchpad.net/ubuntu/+source/mksh/+bug/1058035
was just the beginning, as gcc-snapshot in Debian had it (I
wrote that in the bugreport already), but, now, gcc-4.{6,7}
in sid also have it.

I’ve just tracked it down for gcc-4.6 to have been introdu‐
ced between 4.6.3-9 and 4.6.3-10; gcc-4.7 as currently sits
in wheezy (4.7.1-7) doesn’t have the problem, 4.7.2-2 does.

tg@zigo:~/b $ cat t.c
#include <stdint.h>
typedef int32_t mksh_ari_t;
char ari_sign_32_bit_and_wrap[(

        (mksh_ari_t)(((((mksh_ari_t)1 << 15) << 15) - 1) * 2 + 1) >
        (mksh_ari_t)(((((mksh_ari_t)1 << 15) << 15) - 1) * 2 + 2)

    ) ? 1 : -1];

int main(void) { return (sizeof(ari_sign_32_bit_and_wrap)); }
tg@zigo:~/b $ gcc-4.6 -O -Wextra t.c
t.c:6:62: warning: integer overflow in expression [-Woverflow]
t.c:3:6: error: variably modified ‘ari_sign_32_bit_and_wrap’ at file scope

My question now is, how am I supposed to deal with that? Will
the affected gcc versions transition into wheezy? In that case
I’ll need to apply the hotfix I made for Ubuntu into Debian as
well, so I’d like an answer.

More generally, why does GCC suddenly break such compile-time
check constructs? (mksh guarantees to scripts it runs that all
arithmetic operations are 32-bit and that both signed and un‐
signed operations wrap around, so it *must* know whether the
system does that correctly. And its build script is cross-com‐
pile agnostic, so it doesn’t do run-time checks.)

Thanks in advance,
//mirabilos
-- 
“It is inappropriate to require that a time represented as
 seconds since the Epoch precisely represent the number of
 seconds between the referenced time and the Epoch.”
	-- IEEE Std 1003.1b-1993 (POSIX) Section B.2.2.2