1 |
On 2020-09-08, Neil Bothwick <neil@××××××××××.uk> wrote: |
2 |
> On Tue, 08 Sep 2020 16:12:38 -0500, Matt Connell (Gmail) wrote: |
3 |
> |
4 |
>> > It seems odd that when there's multiple ways to satisfy a |
5 |
>> > virtual there's now way to "configure" which one you want for when |
6 |
>> > that virtual get's pulled in. Maybe I just haven't run into it often |
7 |
>> > enough... |
8 |
>> |
9 |
>> I thought about this myself too. I don't think there's any consistency |
10 |
>> to it either. For example, virtual/jdk will pull in icedtea-bin, |
11 |
>> whereas virtual/rust pulls in dev-lang/rust. It isn't a big deal (to |
12 |
>> me) but I also don't fully understand the motivation for all the |
13 |
>> decisions. |
14 |
>> |
15 |
>> Maybe a thought for a future portage feature setting for virtual |
16 |
>> ebuilds, like --prefer-binary-virtuals versus --prefer-nonbinary- |
17 |
>> virtuals or something? Just spitballing. |
18 |
> |
19 |
> I was thinking along similar lines, or maybe with a prefer-bin USE flag. |
20 |
> |
21 |
> Alternatively, have portage spit out a warning if a virtual is required |
22 |
> and none of its alternatives are installed, but that would spoil running |
23 |
> portage non-interactively. |
24 |
|
25 |
I'd be happy with a simple message when used with --ask that says a |
26 |
virtual package has pulled in "X" by default, but "Y or Z" could also |
27 |
satisfy that virual. |
28 |
|
29 |
If I don't like the default, I can answer "N", do a one-shot install |
30 |
of my preference, and try again. |
31 |
|
32 |
-- |
33 |
Grant |