[Qgis-psc] Handling the Travis CI situation

Denis Rouzaud denis.rouzaud at gmail.com
Fri Nov 6 00:49:17 PST 2020


Le ven. 6 nov. 2020 à 09:13, Matthias Kuhn <matthias at opengis.ch> a écrit :

> Hi all,
>
> Thanks for making the call Nyall and doing the initial work Denis.
>
> A little additional information to put Richard's mail into context.
> The additional work is required even if we roll our own CI instance. It's
> 98% work with respect to dependencies which will also be updated there and
> is not github workflow specific.
>

Yes.
Moving to Github is the part I have done, and it's complete.

As Matthias said, the work required is due to the dependency bump, going
from Ubuntu Bionic to Focal which can be translated as a bump in all our
dependencies (Qt, Python, Proj, GDAL, sip, etc).

For memory, is a list of the (recent) past efforts to our CI/testing
infrastructure:

* introduction of CI for testing (2015?)
* move to Qt5 (2016, https://github.com/qgis/QGIS/pull/2896 and many more
later)
* use Docker within CI to become a bit more agnostic of the CI (2017,
https://github.com/qgis/QGIS/pull/5256)
* move base Docker from Xenial to Bionic (2018,
https://github.com/qgis/QGIS/pull/6355)

Many of these have been done on a voluntary basis.
But the growth of the code base and number of tests, the effort required
today is just too big at once.

Another approach would be to blacklist the failing tests, merge the work as
is and wait for volunteers to revive the tests.
But it's a major downgrade to our quality assurance.

Cheers,
Denis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-psc/attachments/20201106/8d47bb38/attachment.html>


More information about the Qgis-psc mailing list