[QGIS-Developer] PyQgis QgsCoordinateTransform constructor?

[Alessandro Pasotti]

On Tue, Jan 9, 2018 at 4:32 PM, Matthias Kuhn <matthias at opengis.ch> wrote:

> Hi
>
> On 01/09/2018 03:38 PM, Alessandro Pasotti wrote:
> > On Tue, Jan 9, 2018 at 3:14 PM, Andreas Neumann <a.neumann at carto.net
> > <mailto:a.neumann at carto.net>> wrote:
> >
> >     Hi Alessandro,
> >
> >     Thanks for your information.
> >
> >     I just want to hear other core devs' opinions if this would be
> >     useful for QGIS.ORG <http://QGIS.ORG>. QGIS.ORG <http://QGIS.ORG>
> >     infrastructure is already complex and mainly only Jürgen, Richard
> >     and partially Tim understand the infrastructure partially. If we
> >     start something new I would like to see a long-term commitment from
> >     someone to maintain it (lets say at least for the upcoming year from
> >     the start on).
> >
> >
> > I totally agree, I'm not pushing for AWS expecially because the
> > nightlies are still available, just trying to offer a solution to a
> > problem, and the solution is still available (as long as Boundless
> > doesn't switch off the nightlies), when I was in Boundless we went
> > through the same problem 2 years ago, we tried automated docker builds
> > and they timed out (like Denis said) so we went for AWS and it has been
> > working reliably for 2 years, it may just require small adjustments when
> > new dependencies are added (like with Qt 9.1 for 3D).
> >
> > BTW, I believe that a publicly available nightly docker for Python
> > plugins Travis-CI testing is beneficial for everybody and increases
> > plugins quality, I'm just sad when I see people reinventing the wheel.
> >
> > I would be happier if a company committed to sponsor and long-term
> > maintain this part of the infrastructure, today it's up and running and
> > QGIS master nightly docker images for CI are generated and publicly
> > available, but we don't know about tomorrow.
>
> I would much appreciate if we could setup this kind of infrastructure in
> a way that it's maintainable via scripts (in the qgis repo or another
> one). Just like with a .travis.yml file and similar.
>

I think it is already: the whole thing is a bunch of scripts (part of the
docker recipe) and a Vagrant file on a github repo, the purpose of the
vagrant is to build the docker on AWS and push it to docker hub when done.

It's all public so if you want you can just copy it into the main QGIS repo.

But as I said, there is no need to do it now because Boundless is still
building the dockers on AWS every night, unless/until we want to change
something and the PR is not accepted.



> This will allow to spread the load between different developers. E.g.
> when new dependencies are added (see libzip for auxiliary storage and
> similar, qca, ...), we shouldn't need a maintainer to update this dep,
> the one opening the pull request should be able to do this himself.
>
> I guess AWS would be fine for this kind of setup where we only have a
> minimal maintenance effort directly on the infrastructure (as opposed to
> managing physical servers or virtual machines).



Agreed


-- 
Alessandro Pasotti
w3:   www.itopen.it
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20180109/c6ca392e/attachment.html>