------- Additional Comments From joseph at codesourcery dot com 2004-11-07 11:04 ------- Subject: Re: New: mmix-knuth-mmixware testsuite failure: gcc.dg/intmax_t-1.c
On Sun, 7 Nov 2004, hp at gcc dot gnu dot org wrote: > inttypes.h isn't a standard header, neither is it provided by GCC. > The test is wrong and should only be compiled on systems known to > have inttypes.h. (According to notes in PR 448 at least Solaris and > glibc.) That would rather defeat the point of ensuring that GCC's definition of intmax_t agrees with the target's when a target which didn't previously have one acquires one. The use of inttypes.h, a header only required of hosted but not freestanding implementations, is deliberate because of pre-C99 systems that have <inttypes.h> but not <stdint.h>. wchar_t-1.c and wint_t-1.c used to be xfailed on systems lacking <wchar.h>, something similar could be used on systems lacking inttypes.h so when they gain the header the compilation XPASSes but the test for excess errors still fails if the type is incorrect. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18345
