1 |
On Sat, 14 May 2016 00:33:20 +0200 |
2 |
Ulrich Mueller <ulm@g.o> wrote: |
3 |
|
4 |
> >>>>> On Sat, 14 May 2016, Michał Górny wrote: |
5 |
> |
6 |
> > Replace the mentioning of MAKEOPTS configuration variable with |
7 |
> > generic description of 'make options'. This is in line with other |
8 |
> > parts of the specification where we avoid referencing package |
9 |
> > manager configuration explicitly. Furthermore, MAKEOPTS was only |
10 |
> > mentioned this one time with no explanatory description and no |
11 |
> > guarantee of being exported to the ebuild environment, so it is |
12 |
> > better to avoid making developers think it is reliable. |
13 |
> |
14 |
> The council has rejected the EJOBS feature (see bug 273101) twice, |
15 |
> namely for EAPI 5 and EAPI 6. In the discussion about EAPI 6 the |
16 |
> main argument was that similar functionality already exists with |
17 |
> the makeopts_jobs() and makeopts_loadavg() functions provided by |
18 |
> multiprocessing.eclass (see 20140617 council log). Both functions |
19 |
> rely on the MAKEOPTS variable. |
20 |
> |
21 |
> So, IMHO removal of this variable cannot be done in a simple editorial |
22 |
> change but needs to be discussed in the -dev mailing list. We would |
23 |
> also need a replacement for the two above-mentioned functions (which |
24 |
> presumably will require an EAPI bump). |
25 |
|
26 |
Well, the alternative would be to list it properly in ebuild-exported |
27 |
variables section. Considering what you pointed out, this may be |
28 |
a better idea indeed. After all, we re-use it for other build systems |
29 |
as well. |
30 |
|
31 |
-- |
32 |
Best regards, |
33 |
Michał Górny |
34 |
<http://dev.gentoo.org/~mgorny/> |