[Qgis-developer] Quality assurance

[Andreas Neumann]

Agus, I think that developers know about their responsibility. You do
them wrong if you think they regard QGIS as a "testbed".

In my experience, most serious bugs are fixed very quickly. But - you
cannot expect people to work for you all the time for free. When I ask
developers (like Marco and Jürgen) to fix issues, they often do this as
part of paid work (because they developed this new feature) or because I
have a number of support hours I pay them for. After all, for ESRI,
Geomedia, Mapinfo you also pay for support - so why not for QGIS?

In your courses - which version are you using? Also - for macintosh: we
all know that there are many more bugs on OSX. I always tell people
interested in QGIS on OSX that the Windows and Linux version works
better. What can we do if qt (and QGIS) is tested less on OSX?

Regarding testing: can we perhaps partition QGIS in several modules and
assign people to be responsible for QA on a particular module?

As an example, one person may be responsible for testing Postgis data
sources and connections, another person may be responsible for the
raster display, one for projections, digitizing, etc. - if something
seriously breaks in a module the responsible person should make sure it
is fixed before another release.

I for my part would help in such an effort.

Andreas

On 12/17/2011 09:58 AM, Agustin Lobo wrote:
> I agree with Paolo and furthermore think that taking action is urgent:
> qgis has a serious risk of getting the label of "unreliable tool" and
> that label will be very difficult to overcome afterwards.
> I personally confess I feel uncomfortable every time I use it for a
> course because you never know
> what bugs the students are going to find. This last time has worked
> quite well (except on Macs, where
> it has shown to be very unstable), but thanks to the fact that I
> realized the error in georeferencing before the class and
> skipped that exercise. Otherwise, can you imagine the opinion of the
> students towards qgis after
> getting the result shown earlier in this thread? On the other hand,
> styling of rasters is so insufficient that I have to use other
> software to explain the parts related to operations through histogram.
> 
> I personally think that if we are a sufficient number of users willing
> to act as testers we should start doing it right now, not waiting for
> the result of discussions any more.We should consider all releases of
> qgis as RC for us and publish the results of
> our testing as soon as we can after the release. The rest of users
> will decide upon our results.  I understand the interest of many
> developers is having qgis as a test bed and not a producing
> environment and I respect that point of view that makes possible that
> so many improvements and new tools become widely distributed, but
> users requiring a reliable tool must have a tested software with a
> clear summary of the operational problems they will face.
> 
> Agus
> 
> 
> 2011/12/16 Paolo Cavallini <cavallini at faunalia.it>:
>> Il 16/12/2011 18:53, Werner Macho ha scritto:
>>
>>> I dont think we need a Release Candidate cycle - From my point that will
>>> bring only more "confuse" into the release cycle..
>>
>> So why most respected and established projects follow this route?
>> Now QGIS is used very widely in professional contexts, and releasing supposedly
>> stable packages with nasty regressions (as it is happening) brings no good to our
>> users and reputation.
>> I think our QA approach should improve, one way or another.
>> All the best.
>> --
>> Paolo Cavallini - Faunalia
>> www.faunalia.eu
>> Full contact details at www.faunalia.eu/pc
>> _______________________________________________
>> Qgis-developer mailing list
>> Qgis-developer at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/qgis-developer
> _______________________________________________
> Qgis-developer mailing list
> Qgis-developer at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/qgis-developer