Dne 18.2.2014 15:36, Paolo Bonzini napsal(a):
Il 18/02/2014 15:33, Chris Evich ha scritto:
On 02/18/2014 01:28 AM, Lukáš Doktor wrote:
> master. It can happened, but in an odd matter which wouldn't be
> detectable by boot test (eg. windows 2007 installation vs 2003).
Ahh, yes, and IMHO this actually happens much more that you'd believe.
It's why the CI grid testing we do on master is SOOOOOO important.
It's important, and it's not enough. The main benefit of an "automatic
next generator" would be to catch this _before_ it hits master, and
ensure that master is always a good spot for development, without having
to revert buggy series on top.
Paolo
Well no script can guarantee that. Anyway there is another possible
solution:
1) classic review process, when 2 acks are given, instead of applying
the test, we can set label to "READY"
2) every $TIME_UNIT checker would download all READY patches and apply
them on top of master and run extended set of tests. On PASS it would
update the master automatically.
so the merge button would not be used anymore and before update the
master would be extensively tested.
The cons of this solution is, that one nasty "READY" patch could ruin
whole merge. Also it kind-off substitutes the tagged version which Lucas
does for each stable release.
Regards,
Lukáš
PS: Chris, thank you for info about GH features, currently it doesn't
matter how it will be implemented...
It's
unreasonable to ask test/core developers and reviewers to run more than
a quick/simple set. However possible impact to others is big.
_______________________________________________
Virt-test-devel mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/virt-test-devel
