1 |
On Fri, 20 Nov 2015 15:52:32 +0100 |
2 |
Alexis Ballier <aballier@g.o> wrote: |
3 |
|
4 |
> On Fri, 20 Nov 2015 15:26:43 +0100 |
5 |
> Michał Górny <mgorny@g.o> wrote: |
6 |
> |
7 |
> > Dnia 20 listopada 2015 11:19:35 CET, Alexis Ballier |
8 |
> > <aballier@g.o> napisał(a): |
9 |
> > >On Thu, 19 Nov 2015 23:21:52 +0100 |
10 |
> > >Michał Górny <mgorny@g.o> wrote: |
11 |
> > > |
12 |
> > >> And here you removed the newest version having ~ia64 keyword, |
13 |
> > >breaking |
14 |
> > >> revdeps: |
15 |
> > >> |
16 |
> > >> https://qa-reports.gentoo.org/output/gentoo-ci/1061286/7.html#l885 |
17 |
> > >> |
18 |
> > >> I suggest you use eshowkw before removing packages. Please fix or |
19 |
> > >> revert this. |
20 |
> > > |
21 |
> > >done; thx for the reminder |
22 |
> > > |
23 |
> > > |
24 |
> > >feel free to revert such commits (at least mine) and send me an email |
25 |
> > >so that i get notified something went wrong in the future |
26 |
> > > |
27 |
> > >PS: can't these checks/emails be automated ? it must be very boring |
28 |
> > >to manually track the culprit from ci output |
29 |
> > |
30 |
> > Not one that I can think of. This usually requires looking at what |
31 |
> > changed in both packages, and sometimes profiles. I don't want to |
32 |
> > blame wrong people just because someone made an irrelevant commit in |
33 |
> > a package. |
34 |
> |
35 |
> if CI runs after each commit, it should just be a matter of emailing |
36 |
> the author when some new breakage appears |
37 |
|
38 |
If CI were to run after each commit, I'd need a cluster of servers to |
39 |
run it ;-). On the pretty powerful server donated to the task it takes |
40 |
7 minutes for a single run, with 16 parallel jobs. |
41 |
|
42 |
Of course, someone could try to optimize it more, or write a faster |
43 |
checker. I'd really appreciate if someone had the time to do that ;-). |
44 |
|
45 |
At some point, I wanted to make it try 'git bisect'-ing when new |
46 |
failure occurs. However, I don't have the time nor the resources to |
47 |
work on it. In fact, I don't even have time to make it a little bit |
48 |
smarter and figure out which packages are failing. Right now, it only |
49 |
reports changes in number of failures -- if you fix one and break one, |
50 |
it won't report that. |
51 |
|
52 |
So, yes, there's a lot of work and nobody to do it. |
53 |
|
54 |
-- |
55 |
Best regards, |
56 |
Michał Górny |
57 |
<http://dev.gentoo.org/~mgorny/> |