| 1 |
On Sun, Dec 8, 2013 at 12:46 PM, Tom Wijsman <TomWij@g.o> wrote: |
| 2 |
> |
| 3 |
> Given that the retroactive change I suggest causes a lot of complexity; |
| 4 |
> changing it on the next EAPI indeed sounds like one way to go, the |
| 5 |
> alternative is to make it a suggestive guideline or policy and cover |
| 6 |
> it as a QA check in repoman. |
| 7 |
> |
| 8 |
> That QA check could throw a warning when a dependency has no slot. |
| 9 |
|
| 10 |
I think we're better off with defaulting to slot 0 rather than |
| 11 |
erroring/warning if no slot is specified. Otherwise we're basically |
| 12 |
going to have to modify every ebuild in the tree to add the :0 (unless |
| 13 |
there is no warning when only slot 0 exists, but then another slot |
| 14 |
could be added at any time and what is the behavior then?). |
| 15 |
|
| 16 |
Or we could do both - we could define EAPI 6 as having :0 be the |
| 17 |
default slot behavior, and we could have repoman offer soft warnings |
| 18 |
on earlier EAPIs if there is no explicit slot. |
| 19 |
|
| 20 |
(I'd even suggest making the default :0= except that there is no way |
| 21 |
to override that as we defined not depending on a subslot as the |
| 22 |
absence of any operator and not a different operator that is simply |
| 23 |
defaulted.) |
| 24 |
|
| 25 |
Rich |
Archives