1 |
On Fri, 2019-05-24 at 11:00 +0000, Corentin “Nado” Pazdera wrote: |
2 |
> May 24, 2019 11:39 AM, "David Seifert" <soap@g.o> wrote: |
3 |
> |
4 |
> > +case ${EAPI} in |
5 |
> > + [4-7]) ;; |
6 |
> > + *) die "EAPI=${EAPI:-0} is not supported" ;; |
7 |
> > +esac |
8 |
> > + |
9 |
> |
10 |
> Hi, |
11 |
> |
12 |
> I often wondered, why using a eapi-whitelisting logic instead of |
13 |
> eapi-blacklisting? |
14 |
> This kind of change forces to touch the eclass for the next eapi |
15 |
> bump, while being a more defensive |
16 |
> writing style, is it really needed? |
17 |
> |
18 |
> Best regards, |
19 |
> Corentin “Nado” Pazdera |
20 |
> |
21 |
|
22 |
Tbh, there's upsides and downsides to both approaches. Whitelisting |
23 |
means someone has to do some due diligence in checking that everything |
24 |
actually works with a newer EAPI. For most well-written, forward- |
25 |
looking eclasses, they will generally work for newer EAPIs. On the |
26 |
other hand, just look at EAPI 7 and BDEPEND: This required a lot of |
27 |
implicit DEPEND changes (cmake-utils, python-utils, autotools), and as |
28 |
such it was probably for the better that these were EAPI whitelisted. |