1 |
On Sat, 8 Jul 2017 21:05:57 +0200 |
2 |
Ulrich Mueller <ulm@g.o> wrote: |
3 |
|
4 |
> >>>>> On Sat, 8 Jul 2017, Ciaran McCreesh wrote: |
5 |
> |
6 |
> > On Sat, 8 Jul 2017 16:39:29 +0200 |
7 |
> > Alexis Ballier <aballier@g.o> wrote: |
8 |
> >> Indeed, makes sense. Would it also make sense to have some more |
9 |
> >> logical meaning in a future EAPI ? I mean, in every context I've |
10 |
> >> ever seen, applying a rule to the empty set is the neutral of that |
11 |
> >> rule, so that it preserves associativity. |
12 |
> >> That'd mean: || ( ) is false, && ( ) is true, ^^ ( ) is false, |
13 |
> |
14 |
> I have no strong opinion about this. Is it worth the effort of |
15 |
> changing the spec? |
16 |
> |
17 |
> >> ?? ( ) is false. |
18 |
> |
19 |
> I think ?? ( ) aka at-most-one-of should be true if empty. |
20 |
|
21 |
Maybe; this one is annoying I think since it is not associative per |
22 |
definition: |
23 |
?? ( true ?? ( false false ) ) -> ?? ( true true ) -> false |
24 |
?? ( ?? ( true false ) false ) -> ?? ( true false) -> true |
25 |
|
26 |
|
27 |
> > The sensible thing to do is ban it, and additionally ban use? ( ) |
28 |
> > inside || and ^^ (if we've not done that already...). |
29 |
> |
30 |
> Right. We have to check if this will break any eclass generated |
31 |
> dependencies, though. |
32 |
|
33 |
That's probably the best course of action indeed. |
34 |
|
35 |
Alexis. |