| 1 |
El mar, 03-03-2009 a las 11:25 -0800, Zac Medico escribió: |
| 2 |
> |
| 3 |
> You're right. It's easy to says it in words but if you tried to |
| 4 |
> actually implementing you'd find that the PROPERTIES=virtual concept |
| 5 |
> simplifies the problem a lot and makes it possible to optimally |
| 6 |
> solve some cases of bug 141118 that couldn't otherwise be solved |
| 7 |
> optimally without the information that PROPERTIES=virtual provides. |
| 8 |
> |
| 9 |
> What we're talking about is a cost calculation and in order for the |
| 10 |
> cost calculation to be as accurate as possible, we need to know |
| 11 |
> which ebuilds have zero-cost to install (the virtual ones do not |
| 12 |
> install anything directly so they are considered to have zero-cost |
| 13 |
> in themselves). If you consider all ebuilds to have equal cost then |
| 14 |
> your calculation won't be as accurate as it could be. Knowing which |
| 15 |
> ebuilds are virtual gives a hint to the resolver that it should |
| 16 |
> perform a "lookahead" in the cost calculation. |
| 17 |
> |
| 18 |
> We could implement a similar lookahead mechanism for non-virtual |
| 19 |
> ebuilds, but it's more useful for the virtual ones. Backtracking |
| 20 |
> (which isn't implemented yet but is planned for bug 1343) would also |
| 21 |
> be useful for implementing additional cost optimizations (like the |
| 22 |
> one for bug 260225). |
| 23 |
> - -- |
| 24 |
> Thanks, |
| 25 |
> Zac |
| 26 |
|
| 27 |
No problem. Thanks for the explanation :-) |
Archives