1 |
>>>>> On Mon, 18 Jun 2018, Mart Raudsepp wrote: |
2 |
|
3 |
> Ühel kenal päeval, E, 18.06.2018 kell 08:01, kirjutas Ulrich Mueller: |
4 |
>> > > > > > On Sun, 17 Jun 2018, Marty E Plummer wrote: |
5 |
>> > Signed-off-by: Marty E. Plummer <hanetzer@×××××××××.com> |
6 |
>> |
7 |
>> Please don't yet use Signed-off-by. Our policy on this (GLEP 76) is |
8 |
>> still a draft and subject to change. |
9 |
>> |
10 |
>> > -DEPEND=">=sys-apps/sed-4" |
11 |
>> > - |
12 |
>> > +# sed needs to be executable on the build system |
13 |
>> > +BDEPEND=">=sys-apps/sed-4" |
14 |
>> > +[[ ${EAPI:-0} == [0123456] ]] && DEPEND="${BDEPEND}" |
15 |
>> |
16 |
>> These >=sed-4 dependencies are a relic from the time when there |
17 |
>> was a sed-3. It left the tree in 2003 though. |
18 |
>> |
19 |
>> So I suggest to remove this build-time dependency altogether. |
20 |
|
21 |
> I prefer being explicit, but that's probably a lost "battle" for sed, |
22 |
> so, OK, we can remove it instead. |
23 |
|
24 |
sed is guaranteed to be available in the ebuild environment (as well |
25 |
as bash, patch, and find): |
26 |
https://projects.gentoo.org/pms/7/pms.html#x1-11800012.1.1 |
27 |
|
28 |
So in this special case the dependency is entirely redundant. |
29 |
|
30 |
> Otherwise as a generic question for EAPI-7 support in other eclasses: |
31 |
> Doesn't this case leak a BDEPEND variable to ebuilds for lower EAPIs, |
32 |
> as there's no "unset BDEPEND" for the lower EAPIs. And is that OK? |
33 |
|
34 |
IMHO it is not OK. Also "unset BDEPEND" in global scope is not |
35 |
entirely clean, because BDEPEND is just a normal variable in older |
36 |
EAPIs, and could be used by an ebuild. |
37 |
|
38 |
For the eclasses that I have updated, I have done something along |
39 |
these lines: |
40 |
|
41 |
DEPEND="cat-foo/bar" |
42 |
case ${EAPI:-0} in |
43 |
0|1|2|3|4|5|6) ;; |
44 |
*) BDEPEND="${DEPEND}"; DEPEND="" ;; |
45 |
esac |
46 |
|
47 |
It simply assigns DEPEND in older EAPIs, and in EAPI 7 it assigns |
48 |
BDEPEND. The additional DEPEND="" assignment should not harm, because |
49 |
DEPEND is treated specially by the package manager and accumulated |
50 |
across eclasses. |
51 |
|
52 |
Ulrich |