| 1 |
On Sat, 2009-05-16 at 20:21 +0100, Ciaran McCreesh wrote: |
| 2 |
[...] |
| 3 |
> Can't do that. The package manager has to barf on unrecognised .ebuild |
| 4 |
> files. |
| 5 |
|
| 6 |
I assume the reasons are the same as below. |
| 7 |
|
| 8 |
> > If this is not viable, make an unrecognised version string cause the |
| 9 |
> > same fallback as an unsupported EAPI => ignore the ebuild. Again, fast |
| 10 |
> > track to stable portage, and not so long after, you're done. |
| 11 |
> |
| 12 |
> Again, no good. First, it means a year or more wait before doing |
| 13 |
> anything. Second, it removes a whole level of error checking. Third, it |
| 14 |
> means a package manager can't know what versions are available for a |
| 15 |
> package without generating metadata for every potential version. |
| 16 |
|
| 17 |
1) Why a year? I'd say 4-6 months after portage goes stable is fine. |
| 18 |
|
| 19 |
2) Replace the errors with warnings. And these need to exist only at |
| 20 |
'repoman manifest' time, so they're not end-user-visible (and don't need |
| 21 |
to be). |
| 22 |
|
| 23 |
-- Arun |
| 24 |
|
| 25 |
3) This is again, when the metadata is uncached, which is not the normal |
| 26 |
case and can be ignored. And the (minor) performance penalty in the |
| 27 |
cached case, if any, is not reason enough to make this change. |
Archives