[Qgis-psc] Continuous Deployment for OSGeo4W

[Nyall Dawson]

On Mon, 11 Jun 2018 at 22:17, Matthias Kuhn <matthias at opengis.ch> wrote:

> Looking at the core team page of the project [1], there are 3 people involved, 2 of them packaging and one of them taking the big share of things. Thank you Jürgen ;-)
>
> The infrastructure in the background is to my knowledge not documented and we are happy if it just works. And if it breaks if there is someone with knowledge and access rights to fix things.
>
> I just stumbled upon a repository from Oslandia [2] that streamlines the packaging and integrates it with continuous deployment apparently. Having such a repository under QGIS or OSGeo umbrella could help to take away the burden of maintaining the Windows distribution from very few people, having a pull request review scheme as we have for source code and distribute the responsibility.
>
> In short, having a repository where package scripts and build scripts are publicly available, forkable, builds are reproducible and where it's possible to make pull requests.
>
> The main question in the first line is probably, is that something which is considered a QGIS task, an OSGeo task or a task that individual companies should be taking?
>

Thanks Matthias for raising this, Jürgen for keeping things trucking
for so long, and Oslandia for getting on board with this too!

I'm wondering - in the very short term, would it be possible for
someone involved here to bump the GEOS version from 3.5.0 to 3.6.1
(not 3.6.2 -- see below!)? We have a *lot* of crash reports being
filed where the stack trace indicates that the crash is originating
within GEOS, and is hopefully fixed in a more recent GEOS build. From
my following of hub issues I suspect these GEOS related crashes are
the most common crashes experienced by our users on Windows, and we're
seeing an average of 1-2 reports a week with the same origin within
GEOS, yet very different causes within QGIS (e.g. changing symbol
style, creating layout maps, running processing algs, etc).

(I don't think we should go to GEOS 3.6.2 because this regression:
https://trac.osgeo.org/geos/ticket/867 would impact QGIS far too
heavily)

Nyall