[Mitgcm-support] Re: .gnemakerc

[mitgcm-support at dev.mitgcm.org]

Chris Hill wrote:
> We should start including .genmakerc
> files for reference/verification experiments?
> It is useful to know what packages an experiment
> uses.
> 
> In prep. I am changing genmake to
> 
>  o introduce reserved package name of all which allows
> set DISABLE = ( all ) and/or
> set ENABLE  = ( all )
> ENABLE should take precdence over DISABLE.
> Sound OK?
> 
> Chris

I absolutely object.

I don't like disabling by default. What's the sense of disabling mdsio
ever?! Only if you had a replacement which is not the default.

I think all packages should be compilable and enabled by default.
Nothing gets used unless it's called so there's no possible harm.
The only reason AIM is disabled is because it's broken.

I don't see why anyone should be obliged to configure
even more files to make something work. We already have
CPP_OPTIONS.h, CPP_EEOPTIONS.h, data, data.pkg and a data.* for each
real package. .genmakerc is best used for optimizing the model by
changing the default set-up in unique circumstances and I really meant
to delete the ones in aim.* just as soon as AIM was fixed.

If you want to optimize code out then do it on the command line
and subsequently use make makefile.

If you start adding to the list of default DISABLE'd then we
have to start maintaining multiple copies of .genmakerc and it's
already difficult enough to do that for CPP_OPTIONS.h.

If you insist on disabling packages then I'll insist you do this
properly by renaming .genmakerc and combining CPP_OPTIONS.h
CPP_EEOPTIONs.h and .genmakerc into a single config.h which would
be much easier for everyone all around.

A.