http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47642
--- Comment #19 from Jakub Jelinek <jakub at gcc dot gnu.org> 2011-02-14 14:29:12 UTC --- THe output depends. I get: 1.234500e+06 1.234500e+06 333333.3333333333333 333333.3333333333333 1.7 1.7 1.7e+36 1.7e+36 0x1.2d644p+20 0x1.2d644p+20 0X1.4585555555555P+18 0X1.4585555555555P+18 0x1.b333333333333333333333333333p+0 0x1.b333333333333p+0 0X1.476878AP+120 0X1.476878AP+120 1.234500e+06 333333.3333333333333 1.7 1.7e+36 0x1.2d644p+20 0X1.4585555555555P+18 0x1.b333333333333333333333333333p+0 0X1.476878AP+120 for the 23313 testcase and 12340000000000.100000 12340000000000.100000 12,340,000,000,000.100000 12,340,000,000,000.100000 12.340.000.000.000,100000 12.340.000.000.000,100000 12 340 000 000 000,100000 12 340 000 000 000,100000 1.234000e+13 1.234000e+13 ۱٫۲۳۴۰۰۰e+۱۳ ۱٫۲۳۴۰۰۰e+۱۳ for the 23316 testcase. But, the first part of the 23313 testcase relies on printf hook support which is not present on targets other than Linux, so you'll see there something different (e.g. %Qe etc. instead, for odd lines), only the second part which uses quadmath_snprintf should be the same. And the 23316 testcase is a testcase for localization, which you likely don't have either (and is not written to use quadmath_snprintf anyway).