Gentoo Archives: gentoo-dev

From: "Michał Górny" <mgorny@g.o>
To: Andrew Savchenko <bircoph@g.o>
Cc: gentoo-dev@l.g.o, trustees@g.o
Subject: Re: [gentoo-dev] Please subscribe to travis-ci mail alias to get notifications on depgraph breakages
Date: Sun, 12 Apr 2015 18:28:34
Message-Id: 20150412202814.72702c7b@pomiot.lan
In Reply to: Re: [gentoo-dev] Please subscribe to travis-ci mail alias to get notifications on depgraph breakages by Andrew Savchenko
1 Dnia 2015-04-12, o godz. 01:47:43
2 Andrew Savchenko <bircoph@g.o> napisał(a):
3
4 > On Thu, 9 Apr 2015 00:23:29 +0200 Michał Górny wrote:
5 > > Hello, developers.
6 > >
7 > > We have added a new mail alias travis-ci@g.o and set up travis-ci [1]
8 > > to send notifications on status change there. Please seriously consider
9 > > adding yourself to the alias, and contributing to the quality
10 > > of Gentoo.
11 > >
12 > > The mail load is low -- travis will only send notices when the state
13 > > changes from 'good' to 'bad', and the other way around. However, you
14 > > will need to manually grep the logs provided by travis for occurences
15 > > of 'FATAL'.
16 > >
17 > > Please remember that keeping the repository in a broken state is
18 > > inconvenient both for users and other developers. In particular
19 > > the issues include:
20 > >
21 > > 1. dependency resolution errors blocking @world upgrades,
22 > >
23 > > 2. Portage unnecessarily switching packages to ~arch (and therefore
24 > > reducing quality of stable systems),
25 > >
26 > > 3. repoman refusing to commit irrelevant changes to packages.
27 > >
28 > > Therefore, whenever possible please try to fix the issues ASAP.
29 > >
30 > > However, if you are not the person directly responsible for
31 > > the dependency graph breakage, please *reliably* inform him
32 > > about the revert/change you're doing. Failing to do has already
33 > > resulted in developers repeating their mistakes because of
34 > > misinformation.
35 > >
36 > > Preferably, always file a bug in such a case and make it block
37 > > the 'broken-depgraph' tracker [2]. When assigning/CC-ing to the bug,
38 > > please make sure to include both the maintainers of the broken package,
39 > > and the maintainers of the dependency as necessary.
40 >
41 > While I must admit that travis is a quite convenient tool (thought
42 > it has its limitations), I'd like to raise related software freedom
43 > concern.
44 >
45 > Travis itself is a closed, proprietary and non-trivial-to-replace
46 > solution. [...]
47
48 Wrong. Actually, it's trivial. Even better, deploying pcheck
49 on Gentoo-hosted service is likely easier than on Ubuntu container used
50 by travis (though it was quite a nice pkgcore learning experience).
51
52 The only reason I used travis-ci is because they are giving their
53 hardware for free. If someone gave me hardware on sane terms, I can
54 move it elsewhere. Though right now I see no point in wasting Gentoo
55 money on dedicated hardware when the tasks can be offloaded to
56 travis-ci.
57
58 > If travis will change their terms of service in future and our
59 > workflow/infra will depends on these checks, whole development
60 > process may be hampered.
61 >
62 > So developers should think twice before depending their workflow on
63 > this solution. I'm refusing to sign up to the list which in my
64 > opinion indirectly violates Gentoo social contract.
65 >
66 > If some other free tool (preferably deployed on Gentoo
67 > infrastructure) will be used for this task, I'll sign-up right away.
68
69 You can run repoman/pcheck on your hardware too, you know. You don't
70 have to have a dedicated remote service to do that.
71
72 That said, I have bigger plans™ if someone gives me some hardware to
73 handle them. Let's call it 'big Gentoo repository mirror'. The idea is
74 that a script would process repositories.xml, maintain checkouts of all
75 repositories (fetch, sync, remove as necessary), generate md5-cache
76 and provide the resulting repo for syncing via git & rsync. It wouldn't
77 hurt to run travis on all repos there as well.
78
79 In other words, something like our git mirror, though on dedicated
80 hardware and for all repositories :).
81
82 --
83 Best regards,
84 Michał Górny