| 1 |
On Sun, Sep 2, 2012 at 3:02 PM, Michał Górny <mgorny@g.o> wrote: |
| 2 |
> On Sun, 2 Sep 2012 13:54:13 -0400 |
| 3 |
> Mike Gilbert <floppym@g.o> wrote: |
| 4 |
> |
| 5 |
>> On Sun, Sep 2, 2012 at 1:34 PM, Michał Górny <mgorny@g.o> |
| 6 |
>> wrote: |
| 7 |
>> > On Sun, 2 Sep 2012 13:16:38 -0400 |
| 8 |
>> > Mike Gilbert <floppym@g.o> wrote: |
| 9 |
>> > |
| 10 |
>> >> On Sun, Sep 2, 2012 at 8:21 AM, Thomas Sachau <tommy@g.o> |
| 11 |
>> >> wrote: |
| 12 |
>> >> > Mike Gilbert schrieb: |
| 13 |
>> >> >> On Mon, Aug 13, 2012 at 3:38 PM, Dirkjan Ochtman |
| 14 |
>> >> >> <djc@g.o> wrote: |
| 15 |
>> >> >>> On Mon, Aug 13, 2012 at 9:07 PM, Richard Yao <ryao@g.o> |
| 16 |
>> >> >>> wrote: |
| 17 |
>> >> >>>>> Does this seem like a viable option? Is anyone going to |
| 18 |
>> >> >>>>> volunteer to take it from here? |
| 19 |
>> >> >>>> |
| 20 |
>> >> >>>> I can try to make time to make these changes this weekend if |
| 21 |
>> >> >>>> everyone is fine with that. |
| 22 |
>> >> >>> |
| 23 |
>> >> >>> I think we should inform gentoo-dev first. |
| 24 |
>> >> >>> |
| 25 |
>> >> >> |
| 26 |
>> >> >> Just in case someone is waiting for me to take some further |
| 27 |
>> >> >> actions on this, please don't wait on me. I really don't have |
| 28 |
>> >> >> the motivation to carry this plan any further. |
| 29 |
>> >> >> |
| 30 |
>> >> >> If someone else wants to carry the torch, I am happy to assist. |
| 31 |
>> >> >> |
| 32 |
>> >> >> |
| 33 |
>> >> > |
| 34 |
>> >> > Just to get the right summary: |
| 35 |
>> >> > |
| 36 |
>> >> > The plan is now to replace all dependencies in ebuilds, which |
| 37 |
>> >> > have either "dev-lang/python" or something like |
| 38 |
>> >> > ">=dev-lang/python-X.Y" to use "virtual/python" instead? |
| 39 |
>> >> > |
| 40 |
>> >> |
| 41 |
>> >> That's right. |
| 42 |
>> >> |
| 43 |
>> >> > What about versioned dependencies exluding python versions still |
| 44 |
>> >> > in the main tree like ">=dev-lang/python-2.7"? |
| 45 |
>> >> > |
| 46 |
>> >> |
| 47 |
>> >> In this case the virtual would not work. Instead, it could be |
| 48 |
>> >> replaced with this: |
| 49 |
>> >> |
| 50 |
>> >> || ( =dev-lang/python-2.7* =dev-lang/python-3* ) |
| 51 |
>> >> |
| 52 |
>> >> Similarly, >=dev-lang/python-2.6 could be replaced with this: |
| 53 |
>> >> |
| 54 |
>> >> || ( =dev-lang/python-2.7* =dev-lang/python-2.6* |
| 55 |
>> >> =dev-lang/python-3* ) |
| 56 |
>> > |
| 57 |
>> > You could go with virtual/python-2.7, -2.6, -2.5 being an |
| 58 |
>> > appropriate minimal version. |
| 59 |
>> > |
| 60 |
>> |
| 61 |
>> I gave that idea some thought, but I think it would be a somewhat |
| 62 |
>> confusing (it makes my head hurt). |
| 63 |
>> |
| 64 |
>> There are actually a few ways I could see a versioned virtual being |
| 65 |
>> implemented; can you articulate exactly what each virtual version |
| 66 |
>> would have in RDEPEND and how it would be used in other ebuilds? |
| 67 |
> |
| 68 |
> Well, one not-so-crazy idea is that we have: |
| 69 |
> |
| 70 |
> virtual -2.5, SLOT="2.5+": || ( =python-2* =python-3* ) |
| 71 |
> virtual -2.6, SLOT="2.6+": || ( python:2.7 python:2.6 =python-3* ) |
| 72 |
> virtual -2.7, SLOT="2.7+": || ( python:2.7 =python-3* ) |
| 73 |
> |
| 74 |
> # optioanlly, I think it may be nicer than =python-3* in ebuilds |
| 75 |
> virtual -3.1, SLOT="3.1+": =python-3* |
| 76 |
> |
| 77 |
> It's a bit hacky but with it, you can then do: |
| 78 |
> |
| 79 |
> RDEPEND="virtual/python:2.6+" |
| 80 |
> |
| 81 |
> which is quite self-explanatory. |
| 82 |
> |
| 83 |
|
| 84 |
That could work. In this case, we would need to ensure that ebuilds do |
| 85 |
not depend on virtual/python without a slot (or version wildcard). |
Archives