Re: clisp/src makemake.in,1.559,1.560 ChangeLog,1.4687,1.4688

[Bruno H. <br...@cl...>]

Sam wrote:
> > Ignore --with-module=... options for modules that are already
> > contained in BASE_MODULES.
>
> silently ignoring user input is wrong.

We do not ignore relevant input. When he user passes --with-module=regexp,
he means "I want the 'full' linkingset to contain 'regexp'", and makemake.in
implements this.

Before this patch, the user had to know the exact list of BASE_MODULES and
compute the arguments to configure from
   (desired list of modules) AND NOT (list of base modules).
Which most people evidently don't, thus leading to frustrated reports like
this one a few days ago:

>In file included from modules.d:36:
>modules.h:5: conflicting types for `module__regexp__subr_tab'
>modules.h:4: previous declaration of `module__regexp__subr_tab'
>modules.h:6: conflicting types for `module__syscalls__subr_tab'
>modules.h:3: previous declaration of `module__syscalls__subr_tab'
>*** Error code 1

> there should be either a warning
>       "module FOO is in BASE_MODULES, see http://..."

It makes useless work to all those who redistribute clisp with rpm spec files
or similar configuration scripts.

What we can do automatically, without warning, in makemake.in, why warn
about it?

> or the module should be removed from BASE_MODULES and added to MODULES
> (also with a warning).

Nah. BASE_MODULES is fixed. And the user did not mean to say "I don't want
regexp in 'base'". He meant "I want regexp in 'full'".

Bruno