FPU control, round 2

[Paul Brook]

On Thursday 28 April 2005 08:34, François-Xavier Coudert wrote:
> 2. (for gfortran community) Does someone see an objection to removing
> support for GFORTRAN_FPU_ROUND and GFORTRAN_FPU_PRECISION environment
> variables, and add actual code to act upon others (GFORTRAN_FPU_INVALID,
> *_DENORMAL, *_ZERO, *_OVERFLOW, *_UNDERFLOW and *_PRECISION)?

My personal preference would be to remove all the environment variables, and 
replace them with a mechanism for specifying these properties either at 
runtime via a function call, or at link time with magic compiler/linker 
switches.

When this has been discussed before these consensus seems to have been that 
environment variables tend to be more trouble that they're worth. It's too 
easy to have things break unexplainably because an obscure environment 
variable was/wasn't set.

Paul