cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rajani Karuturi <raj...@apache.org>
Subject Re: [PROPOSAL] remove travis pull request checks from github
Date Fri, 30 Oct 2015 06:05:42 GMT
Jenkins though unstable is right at what it does. That is, if its green,
the build works fine, there are no check style errors, there are no new
PMD/Findbugs errors etc.
On the other hand, Travis green/red doesnt say anything at the moment. It
can be red due to a timeout and sometimes its green because it didnt
actually run any tests. showing green without actually testing anything is
dangerous and hence, I +1 disabling it until we can reliably fix it.

@David, Nope they are not related. Jenkins is unstable for reasons we dont
understand and cant see(not related to code but to the jenkins
environment). triggering a "rebuild" there also doesnt work as it builds on
the base branch instead of applying the PR. The only way to get it running
again is force pushing the PR and each run takes more than 3 hrs. sometimes
it requires as many as 4 or 5 force pushes to see it happy and its very
frustrating. Is it possible for some in ASF infra to look at it?

Travis I think is to do with complex configuration we did. But, right now,
no one has the time to look at it and get it fixed. But, this is something
we can fix and is under our control.


~Rajani

On Fri, Oct 30, 2015 at 4:55 AM, David Nalley <david@gnsa.us> wrote:

> 2 hours before this message was sent, there was a message about
> jenkins being unreliable.
>
> So I am starting to think that something other than Travis or Jenkins
> is the problem.
>
> --David
>
> On Thu, Oct 29, 2015 at 5:33 PM, Daan Hoogland <daan.hoogland@gmail.com>
> wrote:
> > Guys and dolls,
> >
> > We have been having travis builds for a while now and they seem to
> generate
> > more work then help. There are a lot of time-outs, (false positive) and
> > when a suite passes it is often because all tests where skipped (false
> > negative). In addition, and this may be just me, I've never been able to
> > pinpoint a problem based on travis output.
> >
> > therefore:
> >
> > I move that we completely disable travis runs on PRs to the cloudstack
> > project. This will reduce confusion about whether a PR is good and will
> > reduce the need for willing contributers to force push their branch over
> > and over.
> >
> > --
> > Daan
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message