CI system maintainability

Christian Mollekopf chrigi_1 at fastmail.fm
Thu Mar 28 19:09:41 GMT 2019


On Thu, Mar 28, 2019, at 7:42 AM, Ben Cooksley wrote:
> Hi all,

Hi Ben,

> 
> Does anyone have any ideas for a long term, proper fix to this?
> 

I believe the proper fix is to make the CI system resilient to such failures.

A project having a broken build for any time period should be "fine", in that it shouldn't result in anyone having to intervene (besides the developers eventually getting around to fix their code).
CI is a tool for developers to help them write better quality code, and some will use it, and some won't, and that shouldn't really matter IMO.

So IMO the question shouldn't be if there should be mandatory code review or not (not that there shouldn't be, I just think it's largely unrelated), but what needs to happen so the CI can just work, no matter if some project chooses to ignore CI errors for a couple of weeks.

I assume a large part of the problem is that if a dependency breaks in CI, users of that dependency are effectively blocked.
To get around this it could for instance make sense to only every use versions as dependency for build that also have passed CI (this is a shot in the dark as I don't really know what the maintainability problem is in the first place).

Cheers,
Christian



More information about the kde-pim mailing list