On Fri, Jan 25, 2013 at 9:16 AM, Thorsten Glaser <[email protected]> wrote: > Daniel Schepler dixit: > >>void test_fn() { >> int i; >> for (i = 1; 0 < i; i *= 2) > > Ah yes, the old GCC “Undefined Behaviour” issue. > > As per a strict reading of the standard, this loop > never terminates because, for signed int, i*2>i is > tau. ☹ > > Adding -fwrapv to the compile command will fix this > (by making signed integer overflow Defined). > > I don’t notice this on MirBSD because we *always* use > -fwrapv there…
And here I went and embarrassed myself by filing #698950... I was wondering why cvs ended up using the old broken AC_FUNC_MKTIME even after I did an autoreconf (after adding an AM_GETTEXT_VERSION or whatever it is to configure.in). On further examination, it looks like the package has an m4/mktime.m4 file with the broken test, so that file should be removed before doing the autoreconf so it can pick up the updated version from autoconf. -- Daniel Schepler -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
