| 1 |
On Mon, 23 Jan 2017 18:17:58 +0100 |
| 2 |
"Paweł Hajdan, Jr." <phajdan.jr@g.o> wrote: |
| 3 |
|
| 4 |
> On 23/01/2017 13:12, Alexis Ballier wrote: |
| 5 |
> > For example, if you allow use.mask or use.force in mixins, you can |
| 6 |
> > end up having unsatisfiable deps that repoman will never catch. |
| 7 |
> |
| 8 |
> Whoa, that sounds bad. Could you elaborate why we wouldn't be able to |
| 9 |
> catch these errors? |
| 10 |
|
| 11 |
Because repoman wont be able to check the 2^{# of mixins} possibilities |
| 12 |
to have something broken there. |
| 13 |
|
| 14 |
> > Arguably, desktop profiles relying on having an useflag forced on a |
| 15 |
> > given package are already semi-broken: they'd be better with the |
| 16 |
> > useflag default enabled and proper usedeps, so the mask/force game |
| 17 |
> > doesnt seem really useful for mixins. |
| 18 |
> |
| 19 |
> Could you give examples of such assumptions? I'd agree in general |
| 20 |
> usedeps sound like the proper solution. |
| 21 |
|
| 22 |
gentoo-x86/profiles/targets/desktop/package.use.force: |
| 23 |
|
| 24 |
# Ensure shared-mime-info is pulled in by glib, otherwise GNOME, XFCE,and |
| 25 |
# numerous gtk-based applications will break, see bug #511894 |
| 26 |
dev-libs/glib mime |
| 27 |
|
| 28 |
> > It'd also be great to have "rules" ensuring all mixins commute, but |
| 29 |
> > I doubt that's easily doable. |
| 30 |
> |
| 31 |
> Could you elaborate more on that, and what the difficulties are? |
| 32 |
|
| 33 |
First you need to define *exactly* what goes into mixins. |
| 34 |
|
| 35 |
Then, you want that a combination of mixins doesnt depend on the order |
| 36 |
they are processed. It is left to the reader to show that it is |
| 37 |
equivalent to that any two mixins commute, that is, the resulting |
| 38 |
profiles with the 'parent' file containing: |
| 39 |
mixinA |
| 40 |
mixinB |
| 41 |
or: |
| 42 |
mixinB |
| 43 |
mixinA |
| 44 |
|
| 45 |
are the same. |
| 46 |
|
| 47 |
If you define mixins having only default enabled or default disabled |
| 48 |
useflags (globally or per package), then commuting is implied |
| 49 |
by 'no two mixins handle intersecting sets of (package,useflag)'. |
| 50 |
|
| 51 |
One can also go the simple way: Defining a mixin priority (either |
| 52 |
explicitly or by alphabetical order for example), so that e.g. in the |
| 53 |
above, mixinA is always processed first. |
| 54 |
|
| 55 |
|
| 56 |
|
| 57 |
Without doing something in that regard I think this is too loosely |
| 58 |
specified and asks for headaches. |
| 59 |
|
| 60 |
Alexis. |
Archives