Re: [linuxtools-dev] Hudson CI NAK'ing patches

[Jeff Johnston <jjohnstn@xxxxxxxxxx>]

I am a bit leery of this.  A test failure should not occur and I'm not
sure everyone will be diligent in tracking their errors.  IMO, tests
that are randomly failing elsewhere should be disabled and fixed by
whoever owns them (i.e. if some non-related test keeps failing, push a patch to
disable it and send a note to the list for someone to fix it).

A verification build can be retriggered after failure if it just hiccups
unexpectedly (e.g. Eclipse.org issue).  If one follows the build started 
message URL and logs in, the build can be retriggered.  A success will
then remove the veto.

Looking at the linuxtools-gerrit job, it appears that Alex, myself, Roland,
Sami, and Patrick have the ability to build.  We could add yourself and others
to the build list.  Worst case scenario is you can ask one of us who has access to
retrigger the build if you feel it is randomly failing.

-- Jeff J.

----- Original Message -----
From: "Alexandre Montplaisir" <alexmonthy@xxxxxxxxxxxx>
To: "Linux Tools developer discussions" <linuxtools-dev@xxxxxxxxxxx>
Sent: Thursday, November 21, 2013 6:01:45 PM
Subject: [linuxtools-dev] Hudson CI NAK'ing patches

Hi all,

What would you think of making the Hudson-Gerrit plugin (a.k.a. Hudson
CI) giving "-1 Code Review" on Gerrit, instead of the current "-1
Verified"? The difference is that a committer can override a -1 CR when
giving it a +2, which allows pushing the patch anyway. When "-1
Verified" is given however, the patch is effectively veto'ed and cannot
be merged at all through Gerrit, unless you rebase it manually, or if
you push it directly through Git.

We do appreciate Hudson CI running the compilation/tests and reporting
the results. It's very helpful! However there are cases where the
failures are unrelated to the code being reviewed (unit test failures in
other projects, random brokage at Eclipse, etc.) and, imo, the human
should be able to ignore what Hudson says. Without having to resort to
dirty tricks like push-it-before-Hudson-runs or pushing through Git.

I'm not sure how to change this, but CDT has their Gerrit set up this
way, if I'm not mistaken.

What do you think?


Cheers,
Alexandre
_______________________________________________
linuxtools-dev mailing list
linuxtools-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/linuxtools-dev