Re: Gfortran and NetBSD 5

[FX <fxcoudert at gmail dot com>]

> 	Some seven thousand test cases fail on NetBSD 5 due
> to "excessive" errors.  These errors are linker warnings.
> The built-in calls to cabs and cabsf in f95_lang.c use the
> compatibility interface.  On NetBSD-5, these function
> should be linked to __c99_cabs and __c99_cabsf not cabs and
> cabsf.
> 	How should this be handled?

Well, if you really want __bultin_cabs to refer to an external symbol  
that has a name other than "cabs", you can look at how it's done for  
darwin in gcc/config/darwin.c and gcc/config/darwin-ppc-ldouble- 
patch.def; look for the PATCH_BUILTIN* macros and the  
darwin_patch_builtin() function.

I hope for the sake of netbsd maintainability that you don't have too  
many such changes to perform (what about other C99 complex functions).  
I'm not an expert at reading standards, but the complex.h  
specification says "The following shall be declared as functions and  
may also be defined as macros. Function prototypes shall be provided."  
and "cabs, cabsf, cabsl - return a complex absolute value". Doesn't  
your cabs() function return a complex absolute value?

FX