1 |
On Thu, 2020-03-19 at 12:17 -0500, Gordon Pettey wrote: |
2 |
> On Thu, Mar 19, 2020 at 9:47 AM Alec Warner <antarus@g.o> wrote: |
3 |
> |
4 |
> > On Thu, Mar 19, 2020 at 6:52 AM Gerion Entrup <gerion.entrup@×××××.de> |
5 |
> > wrote: |
6 |
> > |
7 |
> > > Am Donnerstag, 19. März 2020, 02:59:56 CET schrieb Kent Fredric: |
8 |
> > > > On Wed, 18 Mar 2020 17:52:25 +0000 |
9 |
> > > > James Le Cuirot <chewi@g.o> wrote: |
10 |
> > > > |
11 |
> > > > > Not quite. Tools like repoman will need to be updated to understand |
12 |
> > > > > that an ebuild with KEYWORDS="amd64" can depend on another ebuild with |
13 |
> > > > > only KEYWORDS="noarch". I do think the idea has merit though. |
14 |
> > > > |
15 |
> > > > But the inverse is _not_ true, an ebuild with KEYWORDS="noarch" |
16 |
> > > > *cannot* depend on another ebuild with only KEYWORDS="amd64". |
17 |
> > > Maybe I misunderstand something but shouldn't that be the normal case? |
18 |
> > > Every single Python package (candidates for noarch) for example depends |
19 |
> > > on the Python interpreter, which must have non noarch keywords. |
20 |
> > > |
21 |
> > > |
22 |
> > > > Otherwise it breaks the entire stabilization graph. |
23 |
> > > The condition would be: If the interpreter is stable for an arch, all |
24 |
> > > it's interpreted code is also stable for this arch. |
25 |
> > > |
26 |
> > |
27 |
> > Much of the concern is that this condition is not true for all interpreted |
28 |
> > code. |
29 |
> > |
30 |
> > -A |
31 |
> > |
32 |
> > |
33 |
> > > |
34 |
> > > Best, |
35 |
> > > Gerion |
36 |
> > > |
37 |
> If noarch is an alias that includes all keywords, KEYWORDS="noarch -sparc" |
38 |
> works (in Portage, not sure what PMS says about keyword additivity) and |
39 |
> doesn't break your keyword dependency graph for a Python package that for |
40 |
> whatever reason doesn't function on sparc. |
41 |
> |
42 |
> Semantically, noarch says it hasn't been tested on any arch, so Kent's |
43 |
> evidence follows. Once you get a negative test result from an arch, you add |
44 |
> -arch to that package. It's not called "allarches". |
45 |
|
46 |
So what happens when you try to emerge noarch package that has -sparc |
47 |
dependency on sparc? Right now, the package manager is able to |
48 |
immediately tell you it's wrong. With noarch, for every package you |
49 |
have to traverse the whole dependency graph to make sure. |
50 |
|
51 |
Read: things will become much slower. It's not like they're fast |
52 |
anyway. |
53 |
|
54 |
-- |
55 |
Best regards, |
56 |
Michał Górny |