[Qgis-developer] [Fwd: Re: [Qgis-user] Poll on website]

Martin Dobias wonder.sk at gmail.com
Wed Jul 15 16:24:42 EDT 2009


Hi Tim

On Wed, Jul 15, 2009 at 9:02 PM, Mailing Lists<lists at linfiniti.com> wrote:
> Hi
>
> On Wed, Jul 15, 2009 at 7:44 PM, Martin Dobias<wonder.sk at gmail.com> wrote:
>
> 8<-------------snip-------------------
>
>>
>> My biggest concern is about the naming: calling 1.x versions
>> "unstable" seems unfair to me. Any greater features are developed and
>> tested in branches and merged once they're ready. So the trunk is +-
>> stable and virtually ready for a release at any time. That's why I
>> would prefer to mark also 1.x releases "stable" or avoid
>> stable/unstable completely. Every 1.x release brings many new features
>> that make 1.0 look old :-)
>
> Yes developing in branches helps keeps trunk more or less release
> ready but introducing new features inevitably will introduce new bugs
> / crash points. Although we have tried to introduce a culture of unit
> testing, this as not reallly taken hold, and that means we put the
> burden of testing and finding those bugs and crash points on our
> users.

I'm also sorry for the failure of creation of unit tests for qgis. I'd
be glad to discuss it on the hackfest in Vienna. There are some
possibilities for improvement. Main reason I see is that it's hard to
motivate anyone to create unit tests for older code which just works
(tm) or mostly seems to work...

> If we are going to do that it seems to me only fair that we
> indicate this with the 'unstable' moniker.

I still think that stating a product is "unstable" makes bad marketing
impression. This would mean that any project should release their new
versions as unstable. Moreover I feel that our stable and unstable
releases are pretty equal when it comes to stability.

> There is a good quote that
> says 'software without unit tests is broken by design'....

Well, that's maybe too tough :-)
Anyway, the good news is that symbology-ng includes a set of unit tests!


>>>> And I think that's already the plan.  1.2 will be released when the
>>>> symbology-ng & labeling stuff was merged.
>>>
>>> Why don't we release another version *before* this? It is possible that
>>> major merges will cause some instability, therefore pushing forward 1.2.
>>
>> Recently I've been wondering what will be the best time for new
>> release (1.2). In about two weeks we're going to finish the
>> qgis-mapper project and there are some more functionality besides new
>> symbology and labeling I'd like to put into trunk: most importantly
>> OpenStreetMap provider+plugin, topology checking plugin and vector
>> features caching. But that's probably too much for one release... So
>> one possibility would be to release 1.2 with just some of the features
>> and get the rest into 1.3. Opinions?
>
> If all the new features are ready, I'd vote to introduce them all in
> the 1.2 release. If Martin prefers, we can do a release of 1.2 now as
> suggested above and then merge in his new stuff and do a 1.3 release
> in the not too distant future....

I'll leave that decision on you, Tim :-)
Basically the plugins shouldn't do any harm to application itself. The
only risky thing would be the patch for caching of vector features -
it gives a boost to the vector rendering speed in expense of some
memory.
Regarding labeling - I'm not sure whether to release it as a
standalone (c++) plugin and only later put it into core library - or
put it directly to the core library. Anyway we have to keep the old
labeling code due the api stability and just allow the user to switch
between the old/new implementation - the same will apply for the
symbology.

Martin


More information about the Qgis-developer mailing list