Dear R developers I wrote a package whose computational 'engine' is programmed in Fortran 77. However, I used also C code, mainly for exploiting the Mathlib Standalone set of functions. I successfully compiled and used my package for months, under R-2.4.1 (and MINGW - GCC 3.4.2). Now, I switched to R-2.6.0 (and the 'Rtool.exe' set of Murdoch, installed EXACTLY as in its prescriptions) and I encountered problems during compilation. In a C function I use the DOUBLE_EPS constant (of course, I loaded the <R_ext/Constants.h> header). but the the 'R CMD check' say that " CFunctions1.c:333: error: `DBL_EPSILON' undeclared (first use in this function) ... " I suspect that the difference between my two compilation experiences lies in the different structure of the <float.h> header between GCC 3.4.2 and GCC 3.4.5. I'm not a C expertise and hence the question: There is a clean way for using this kind of constants (DOUBLE_EPS and other similar) as my previous experience? Thanks a lot in advance Fabrizio P.S. I have a Pentium Centrino and Windows XP.