1 |
>>>>> On Sun, 8 Jun 2014, Rich Freeman wrote: |
2 |
|
3 |
>> b) failglob in global scope |
4 |
>> Bug #463822 |
5 |
|
6 |
> I'm a bit concerned that this one could be a cure worse than the |
7 |
> disease - unintended side-effects/etc. A repoman check seems a lot |
8 |
> less likely to cause problems. |
9 |
|
10 |
Any specific example of a side effect that you have in mind? I believe |
11 |
that there are none in global scope, but maybe we shouldn't extend the |
12 |
feature to function scope. |
13 |
|
14 |
>> 4. Features rejected from EAPI 5 |
15 |
|
16 |
> Ok, I suspect we're going to spend about 95% of the time on these. |
17 |
|
18 |
Yes. ;-) |
19 |
|
20 |
>> a) Patch applying function in package manager |
21 |
>> Bug #463768 |
22 |
>> - Needed for 2d) and 4b) |
23 |
>> - This will duplicate epatch() from eutils, in simplified form. |
24 |
>> - Name "eapply" has been suggested. |
25 |
|
26 |
> My two cents. Keep an eclass for the fancy stuff, but move into the |
27 |
> EAPI a few basic patch functions: |
28 |
|
29 |
> 1. Application of a patch file, defaulting to -p1, allowing override |
30 |
> of -p level, but without auto-detection. |
31 |
> 2. Application of a directory full of patch files in lexical order, |
32 |
> again supporting override of -p level without auto-detection. |
33 |
|
34 |
> I'm not a big fan of either auto-detection or forcing patch re-writes |
35 |
> to be -p1 always. |
36 |
|
37 |
This more or less agrees with the status of the discussion: |
38 |
https://bugs.gentoo.org/show_bug.cgi?id=463768#c32 |
39 |
|
40 |
We haven't decided yet if there should be autodetection, but I'd also |
41 |
prefer a -p1 default without autodetection. There's always epatch if |
42 |
you need anything more fancy. |
43 |
|
44 |
>> b) User patches |
45 |
>> Bug #475288 |
46 |
>> - Needs 4a) |
47 |
>> - Current wording of the spec requires that every ebuild must |
48 |
>> include a call to the function in src_prepare, which is |
49 |
>> controversial. |
50 |
>> - Names "apply_user_patches" or "eapply_user" have been suggested. |
51 |
|
52 |
> I'm generally fine with it (we can bikeshed the name to death later), |
53 |
> but is there any reason to not just simply apply the patches after |
54 |
> src_prepare is done and not make maintainers call it? If we're |
55 |
> concerned about sequencing with other activities in src_prepare then |
56 |
> that is fine - I don't have a big problem with it being a fatal error |
57 |
> if it doesn't get called. |
58 |
|
59 |
IMHO the package manager shouldn't perform any operations between |
60 |
src_* phases, outside of phase functions. Therefore, the user patch |
61 |
function should be called either from src_prepare, or from a phase |
62 |
function (src_userpatch?) newly introduced. |
63 |
|
64 |
I'd strongly dislike a new phase function, but conceptionally, it |
65 |
would still be much cleaner than doing some magic after src_prepare. |
66 |
|
67 |
My suggestion would be to add the user patch function to |
68 |
default_src_prepare, plus a policy that the function must be called if |
69 |
an ebuild (or eclass) contains an explicit src_prepare(). The latter |
70 |
could be enforced by repoman. |
71 |
|
72 |
>> c) EJOBS variable |
73 |
>> Bug #273101 |
74 |
>> - Discussion was in 2008. Is there (still) consensus? |
75 |
|
76 |
> Assuming anybody still wants to implement this, I see no reason to |
77 |
> keep bikeshedding this. EJOBS fits the bill. |
78 |
|
79 |
> The only thing that might be worth noting is that distcc users may |
80 |
> have an interest in distinguishing between gcc jobs and everything |
81 |
> else. I once messed with dictcc with very high job numbers and it |
82 |
> worked great when make hit a directory full of .c files, and not so |
83 |
> great when make/ant/whatever tried to run 50 instances of java in |
84 |
> parallel. |
85 |
|
86 |
I don't have a strong opinion about this one. |
87 |
|
88 |
However, my general advice would be not to impose work on PMS and |
89 |
package manager teams, for features that nobody is pushing any longer. |
90 |
|
91 |
>> d) Source eclasses only once |
92 |
>> Bug #422533 |
93 |
|
94 |
> Does anybody still want this included? It seems to me like the list |
95 |
> discussion was leading in a different direction, but it isn't 100% |
96 |
> clear to me if this is the case. |
97 |
|
98 |
Right, discussion has died down. Many eclasses now use a mechanism |
99 |
equivalent to the one in C header files, which seems to do the job. |
100 |
|
101 |
>> e) HDEPEND: host dependencies for cross-compilation |
102 |
>> Bug #317337 |
103 |
|
104 |
> It isn't 100% clear to me which of the 14 proposed solutions in that |
105 |
> bug is the final answer, but I think it is DEPEND becomes target |
106 |
> depends, and HDEPEND is introduced as host depends. |
107 |
|
108 |
> Is the final intent that if HDEPEND isn't defined, then |
109 |
> HDEPEND=DEPEND, or does HDEPEND=""? |
110 |
|
111 |
We did RDEPEND=${DEPEND} once and it was a mistake. So I think default |
112 |
HDEPEND should be empty. |
113 |
|
114 |
Ulrich |