1 |
On Thu, 2003-11-20 at 15:23, Aron Griffis wrote: |
2 |
> Spider wrote: [Wed Nov 19 2003, 11:48:20AM EST] |
3 |
> > Don't emake and econf both fail if they fail, making || die "" |
4 |
> > unnecessary and even impossible? |
5 |
> |
6 |
> It's really poor practice to depend on this. I argued against this |
7 |
> change and was ignored. Here are some reasons: |
8 |
> |
9 |
> - Calling die from econf/emake defeats the function and line number |
10 |
> reporting that are part of die. |
11 |
> |
12 |
> - Developers have to keep track of which ebuild.sh functions call |
13 |
> die and which ones don't. It's really better to leave the error |
14 |
> handling in the ebuild in all cases so that the practice is simply |
15 |
> to call die in all situations where it is appropriate. |
16 |
|
17 |
I agree. I asked seemant if we could make that official policy: |
18 |
15:37 <@seemant> g2boojum: I agree 200% with that |
19 |
15:37 <@seemant> the die *should* be in ebuilds |
20 |
|
21 |
If nobody complains too loudly, can we get this policy into the docs? |
22 |
|
23 |
-g2boojum- |
24 |
-- |
25 |
Grant Goodyear <g2boojum@g.o> |