1 |
On Wednesday, June 22, 2011 11:16:25 justin wrote: |
2 |
> On 6/22/11 4:47 PM, Andreas K. Huettel wrote: |
3 |
> > Am Mittwoch 22 Juni 2011, 16:35:07 schrieb Matthew Summers: |
4 |
> >> One thing to note is that a few various python modules, like numpy, |
5 |
> >> really benefit from fortran. While many people using numpy are |
6 |
> >> scientists, many are not and further there are various modules that |
7 |
> >> depend on numpy that non-science folks use. I, for one, care little |
8 |
> >> about where that flag is set, since I have manually set that USE for |
9 |
> >> years now in make.conf. I am simply hoping to make you aware of the |
10 |
> >> fact that there are potential cases that could be easily overlooked. |
11 |
> > |
12 |
> > ... and there are also typical end user software packages like digikam |
13 |
> > and kipi-plugins for photo processing, which rely e.g on opencv for face |
14 |
> > recognition, red eye removal, ... |
15 |
> |
16 |
> So then, |
17 |
> |
18 |
> lets make a clear decision how we should handle the fortran USE. |
19 |
> |
20 |
> 1. leave it as it is as global USE enabled. |
21 |
> |
22 |
> 2. enable it as default for gcc |
23 |
> |
24 |
> 3. let package maintainers fix their packages, |
25 |
> means depend on fortran by default where it is the best and/or needed |
26 |
|
27 |
(3) needs to be done regardless of anything else. default settings doesnt |
28 |
prevent people from disabling fortran on their systems and then hitting a |
29 |
broken build. |
30 |
|
31 |
if (3) is done, does it really need to be a global flag anymore ? |
32 |
-mike |