| 1 |
On Tue, 30 May 2017 09:56:07 +0100 |
| 2 |
Ciaran McCreesh <ciaran.mccreesh@××××××××××.com> wrote: |
| 3 |
|
| 4 |
> First problem: encoding "don't change this from its current setting |
| 5 |
> unless you have a reason to do so" is an utter pain in SAT. |
| 6 |
|
| 7 |
I get the impression that this is harder to solve in Gentoo than it has |
| 8 |
to be, because my impression of portage config is that all sources of |
| 9 |
configuration are equal. |
| 10 |
|
| 11 |
Configuration |
| 12 |
in /etc/portage/* , /etc/portage/profile/*, /usr/portage/profile/* are |
| 13 |
taken like equals, and as far as I can tell, portage has no way of |
| 14 |
knowing if a use flag is disabled by user choice, or if the flag being |
| 15 |
disabled is simply the default. |
| 16 |
|
| 17 |
I think having a clear way to disambiguate between the two would give |
| 18 |
the resolver more freedom to adjust the settings that the user didn't |
| 19 |
specify as preferences, and try harder to conform to the users |
| 20 |
preferences.... |
| 21 |
|
| 22 |
Instead of the current approach, where you can either "Change nothing" |
| 23 |
or "change anything". |
| 24 |
|
| 25 |
"Change nothing at first" -> "Change anything within this subset |
| 26 |
second" -> "Recommend changes in config if you can't solve within that |
| 27 |
constraint" |
Archives