Gentoo Archives: gentoo-dev

From: Pacho Ramos <pacho@g.o>
To: gentoo-dev@l.g.o
Subject: Re: [gentoo-dev] News for Gentoo CI and pull request CI
Date: Sat, 12 Dec 2015 09:46:20
Message-Id: 1449913564.15797.24.camel@gentoo.org
In Reply to: [gentoo-dev] News for Gentoo CI and pull request CI by "Michał Górny"
1 El vie, 11-12-2015 a las 23:13 +0100, Michał Górny escribió:
2 > Hi, everyone.
3 >
4 > I'd like to make a short announcement that in the last few days I've
5 > been using my little free time to hack on my CI scripts, and they had
6 > a few major improvements. Most notably:
7 >
8 > 1. pkgcheck now checks for outdated Manifests. In other words,
9 > gentoo-ci will most likely complain about missing or extraneous
10 > Manifest entries before Infra does, and pull request checks should
11 > complain loudly about them before the PRs are merged.
12 >
13 > 2. The scripts use XML output of pkgcheck and group issues by
14 > packages,
15 > rather than grepping text output for failure indications
16 > and considering successive lines a single failure. Which means better
17 > output, better permalinks, better navigation and a completely new
18 > ugly
19 > UI.
20 >
21 > 3. Most of the persisting issues (like deprecations) are skipped now,
22 > and all the output fits on one page. And that page is usually easier
23 > on
24 > the browser than each of the splits were.
25 >
26 > 4. gentoo-ci compares previous check results correctly, and splits
27 > the output into three groups: new issues, previous issues and fixes
28 > issues.
29 >
30 > 5. gentoo-ci bisects on new issues, pinpoints the exact commit
31 > introducing the failure and CCs both the author and the committer.
32 > In other words, does the hard work for me and does it more
33 > efficiently.
34 > It also has nice bisect caching support which makes bisecting a
35 > number
36 > of issues caused by the same commit very fast.
37 >
38 > 6. pull-request checks now distinguish between issues introduced by
39 > PR
40 > and carried over from ::gentoo. Not that you should ignore the latter
41 > entirely since they may hide additional issues from PR.
42 >
43 > Most of this fun stuff came from a simple obvious thing that only
44 > recently occurred to me: I don't have to run a full pkgcheck for
45 > bisecting! I just need to check the packages that are failing
46 > in the most recent commit!
47 >
48 > Of course, the whole thing is open source:
49 >
50 > 1. https://bitbucket.org/mgorny/repo-mirror-ci
51 >
52 > 2. https://bitbucket.org/mgorny/pkgcheck-result-parser
53 >
54 > Enjoy!
55 >
56
57 Thanks a lot!