Re: [jamboost] Re: [BBv2] warning support

From: Vladimir Prus (ghost_at_[hidden])
Date: 2005-09-19 04:27:46

On Saturday 17 September 2005 22:09, Reece Dunn wrote:
> Andrey Melnikov wrote:
> > Reece Dunn wrote:
> >>Alexey Pakhunov wrote:
> >>>Also the feature should not be 'propagated'. It should be 'free' because
> >>>it does not affect the result of compilation. It only affects the
> >>>warnings that will be displayed.
> >>
> >>Thanks for the clarification... I will update this :).
> >
> > According to http://tinyurl.com/7grs8#bbv2.reference.features.attributes
> > it isn't free either, because it does have a limited set or values. IMO
> > it's "incidental".
>
> Thanks for the link! In the description of incidental: "A feature that
> controls a compiler's warning level is one example of a likely
> incidental feature." :). Even Volodya agrees with you.

IIRC, that sentence text belongs to Dave ;-) But yes, if you omit "incidental" 
you'll get warning-as-error-true in the target path, which is not what you 
want.

> I have provided an updated diff based on feedback.

> +feature warnings :
> + on # turn warnings on

Heh, what does that mean? Which warnings are turned on? I'd prefer this
to be spelled down.

> + off # turn warnings off

All warnings? Again, I'd rather have this spelled down.

> + all # enable all warnings

> +flags gcc.compile CFLAGS <warnings>on : ;

This can be removed, as it has no effect.

> +flags gcc.compile CFLAGS <warnings>off : -w ;
> +flags gcc.compile CFLAGS <warnings>all : -Wall ;

I'd argue that <warnings>all should be a default. Noted that gcc.jam now has a 
hardcoded -Wall, so:

1. You proabably need to remove that hardcode, or verify that feature indeed 
overrides -Wall (I think so).
2. If you remove hardcode, I'd prefer <warnings>all to be default.

- Volodya

-- 
Vladimir Prus
http://vladimir_prus.blogspot.com
Boost.Build V2: http://boost.org/boost-build2