1 |
On Tue, 09 Sep 2008 20:45:52 +0400 |
2 |
Peter Volkov <pva@g.o> wrote: |
3 |
> В Пнд, 08/09/2008 в 23:34 +0000, Jorge Manuel B. S. Vicetto пишет: |
4 |
> > So we're talking about adding the following to EAPI-2: |
5 |
> |
6 |
> While it's not too late. Can we make dobin, doman and other do* |
7 |
> functions finally die in EAPI=2? I've reviewed discussions on -dev |
8 |
> [1],[2] and bug 138792 [3] and seems that the only possible stopper is |
9 |
> that implementing them as functions makes impossible to use them with |
10 |
> xargs. Maybe for such rather rare case we should create new functions |
11 |
> (xdo{bin,*} or whatever name is better)? |
12 |
|
13 |
I'd suggest holding off on that one. There're at least three different |
14 |
ways of implementing it, all with different implications, and it needs |
15 |
proper discussion. |
16 |
|
17 |
* Using traps looks nice on the surface, but in practice they're |
18 |
sufficiently weird on things like conditionals that they're probably not |
19 |
a useful solution. |
20 |
|
21 |
* Banning xargs and doing them as functions is a possibility, but far |
22 |
from ideal, especially since it's just working around a Portage |
23 |
limitation. |
24 |
|
25 |
* Making Portage support subprocess dies is the nice solution, but this |
26 |
probably isn't an EAPI 2 timeframe feature. |
27 |
|
28 |
In addition, having nonfatal versions of commands is also useful in |
29 |
practice. Exheres has a 'nonfatal' command, so you can do 'nonfatal |
30 |
dodoc foo bar baz'. This also needs discussing before deciding upon a |
31 |
spec. |
32 |
|
33 |
-- |
34 |
Ciaran McCreesh |