Re: [PATCH] More %define/%code encapsulation

[Paolo Bonzini]

> I fail to see how this:
>
>   b4_get_percent_code_ifval([[package]], [.])
>
> is better than this:
>
>   m4_ifval(b4_percent_code_get([[package]]), [.])

For one thing, it's an exception to the rule that m4 macros almost 
always accept quoted material.

> In my opinion, if you want to make life easier for skeleton authors, give 
> them less macros to understand.  If they can't grasp how to compose small 
> concepts into larger ones, they're in the wrong business, and they should 
> at least stay away from any UNIX-like OS.  As far as M4 quoting, they're 
> going to have to understand it anyway if they're working with Bison's 
> skeletons.

Yes, but we can at least provide something more consistent.

> If we try to specialize every m4_FOO for every Bison 
> construct, we're going to gradually create a large mess in bison.m4.  

There is no other Bison construct to specialize for.  In the case of 
_code, for example, only _ifdef may make sense; but I'd be more inclined 
to drop that one as it is unused, rather than all the b4_percent_define 
macros.

Previous to your reorganization of %define, it defined macros directly. 
 While bad, this at least was relatively terse: m4_ifval([b4_package], 
[.]).  I would like to rephrase this as 
b4_percent_define_ifval([package], [.]).

> Since none of this appears to be documented, why don't we just drop the 
> no-value form?

I kind of like `%define public', much more than `%define public "true"'. 
 So I think it'd be a pity to remove it, but on the other hand I have 
no big objections.  However, I would like to hear from Akim since he 
introduced it last July.

Paolo