[QGIS-Developer] Some thought on LTR

[Matthias Kuhn]

Hello list,

Some more thoughts on this topic, because it comes up repeatedly.

One the first principles of every of our developers is to try to avoid 
regressions. Whenever a QGIS developer writes code or reviews code, 
there is this process in the back of the head going on thinking "could 
this cause side effects?". Most of the times we are quite efficient in 
detecting side effects and if in doubt running some tests (manually or 
by adding unit tests) to check and adjust if required. This is what 
makes QGIS a rock solid piece of software (are you in doubt? Take QGIS 
2.0 and QGIS 3.4 and perform some common GIS tasks). It's pretty awesome 
actually!

Once in a while though, this process fails to detect a risk or 
implication or use case of which "the developer" was not aware [1]. 
Unfortunately we are not able to completely avoid this situation.

We could of course

   a) Throw more review power to it. If we have the resources to 
download, build and run every pull request, that will help to catch some 
additional issues.

   b) Perform more real world testing. The more testing is done on 
nightly builds, the quicker the turnaround to detect issues.

   c) Combined with b) have more continuous developer power for 
bugfixing (not only before a release)

   d) Stop shipping bugfixes (hint: joke, that makes the LTR concept 
pointless and anyone can do that today already by sticking to the .0 
patch release ;) )


As you also, Matteo, I am not able to offer a solution (unfortunately). 
So this is mostly food for thoughts.

Any initiative or strategic decisions to improve a), b) and c) is very 
welcome.


At the same time, let's not obstruct our view too much by outliers. In 
my opinion we are doing very well with what we have available and are 
overall heading in a good direction!


Let's rock on and make QGIS even better

Matthias


[1] A slight exaggeration, but nice illustration nevertheless 
https://xkcd.com/1172/


On 8/2/19 8:46 AM, Matthias Kuhn wrote:
> Hi Matteo
>
> On 8/2/19 7:15 AM, matteo wrote:
>> Hi,
>>
>>> Right -- but again, I see no concrete issues here. We have one issue
>>> which isn't a regression (postgis rasters) and is instead a feature
>>> request, and one regression which was caused by an important fix
>>> (broken grass for Windows non-English users), not a feature! (And one
>>> unknown issue cos you pasted the wrong link ;)
>> just to finish the thread:
>>
>> https://github.com/qgis/QGIS/issues/30787
>
> Same situation as with grass here:
>
> - A first issue was reported
>
> - The fix for the first issue had unintended and unlucky side effects 
> (a risk that comes with every bugfix/change unfortunately)
>
> - Followup problem was reported
>
> - Patch was supplied within some days after reporting the regression
>
> Matthias
>
>>
>>> Also - just to point out - Alessandro implemented the feature request
>>> overnight. Big kudos to him -- we finally have a stable way of using
>>> PostGIS rasters in QGIS.*
>> kudos to Alessandro and thanks for sharing your thoughts
>>
>> Matteo
>> _______________________________________________
>> QGIS-Developer mailing list
>> QGIS-Developer at lists.osgeo.org
>> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>>
> _______________________________________________
> QGIS-Developer mailing list
> QGIS-Developer at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>