See the following errors: # gcc -o test -D_GNU_SOURCE test.c In file included from /usr/include/stdio.h:933:0, from test.c:1: /usr/include/bits/stdio-ldbl.h:67:1: error: 'vdprintf' undeclared here (not in a function) /usr/include/bits/stdio-ldbl.h:68:1: error: 'dprintf' undeclared here (not in a function) test.c: In function 'main': test.c:9:16: error: called object 'dprintf' is not a function Seems to be related to in appropriate ifdef. Full test case below: https://gist.github.com/1927104
The report omits the essential information, which is the configuration. This only affects __LDBL_COMPAT configurations such as powerpc.
I committed a fix on the trunk (commit 36c8acb) though I did not test it on an affected configuration. I don't think the bug actually appeared for this case on the trunk beforehand, because _GNU_SOURCE gets __USE_XOPEN2K8. But probably there was some other bug in 2.13 like _POSIX_C_SOURCE defaulting wrongly for _GNU_SOURCE.
Rebuilding gcc with 'with-long-double-128' also fixes this issue... that makes sense though, since these DEFINES are sort of related to that bit? Going to retest with just the header fix now. -M
Yes, that's expected. The problematic code is for machines that can have a new long double type, but previously didn't.