Re: [PATCH 0/4] Fixes for the --without-default-features configure switch

[Paolo Bonzini]

On 13/07/21 16:54, Cole Robinson wrote:
> Patches look fine and fix some issues but others persist
> (--disable-system isn't triggered).

I wouldn't say --disable-system counts as a feature, since it's really a 
shortcut for choosing a subset of the targets.  Likewise for linux_user 
and bsd_user.

> IMO this needs an audit, but more
> importantly 'configure' should be rearranged a bit to make this less
> likely to regress:
>
> * move all the --enable/--disable variable init into one section with
> nothing else mixed in
>
> * convert the values to all use
> $default_yes/no/auto/enabled/disabled/... variable syntax so visually
> it's consistent, and if a default is ever changed like $default_no ->
> $default_yes then we behave correctly (as opposed to 'no' -> 'yes').

This is a nice idea.  We should only have default_yes/no/auto, plus 
"auto" for Meson options.

Also there's the idea of parsing --enable/--disable options for Meson 
options automatically from the introspection data.  This has the 
advantage that you get the default automatically from meson_options.txt 
and -Dauto_features, without any code in configure.

Paolo