James Youngman wrote: > > <float.h> exists on all systems. But on BeOS the LDBL_* macros are wrong, > > which causes gnulib's isnanl() to consider some finite values as being NaN, > > which in turn breaks the *printf routines. > > This is useful information. Do you think it might be worthwhile to > include it as a comment somewhere in the software, too? There are > very many gnulib modules. The reader will frequently need to ask "why > might I need to use this module?".
You are right. Users want to know the answer to the question "why need I gnulib if all I need is <float.h>?". The answer to this question is answered in the documentation (not somewhere in comments). Look at doc/headers/float.texi. Bruno
