float.h is missing mandatory C11 fp macros like DBL_DECIMAL_DIG and LDBL_DECIMAL_DIG #26657
Description
| Bugzilla Link | 26283 |
| Version | unspecified |
| OS | Linux |
| Reporter | LLVM Bugzilla Contributor |
| CC | @hfinkel,@itaraban |
Extended Description
Hi, using Clang 3.8 from the Ubuntu ppa I found that float.h does not define some mandatory C11 floating point constants (as per N1570 draft: 5.2.4.2.2p11).
Steps to reproduce: run this simple diagnostic snippet, where Clang fails with "error: use of undeclared identifier 'LDBL_DECIMAL_DIG'", despite the proper include.
#include <stdio.h>
#include <float.h>
int main(void) {
long double x=0.3L;
printf("v=%s; x=%.*Lg\n",VERSION,LDBL_DECIMAL_DIG,x);
return 0;}
Expected: the value is to be printed with the right decimal places, as defined by the C11 spec.
I have checked that at least DBL_DECIMAL_DIG and LDBL_DECIMAL_DIG are not defined on /usr/lib/llvm-3.8/lib/clang/3.8.0/include/float.h, but internal versions with underscores (like LDBL_DECIMAL_DIG) seem to exist elsewhere, but are not exposed in a standard compliant way. I can't say if the underscored versions are "correct" in that they evaluate to the proper value for all target architectures.
All versions of Clang ought to be affected, but only 3.6 and 3.8 were tested. A possibly (vaguely) relevant mailing list post exists: http://lists.llvm.org/pipermail/cfe-commits/Week-of-Mon-20150209/123212.html
GCC/glibc work correctly as expected since at least 4.8.4, but possibly sooner.
No other fp macros were tested, but the above ML post suggests that build tests that check for them are not universally available, so more macros might be missing.
Output of: clang-3.8 --version
Ubuntu clang version 3.8.0-svn257311-1~exp1 (trunk) (based on LLVM 3.8.0)
Target: x86_64-pc-linux-gnu
Thread model: posix
InstalledDir: /usr/bin