CI Requirements - Lessons Not Learnt?

Martin Gräßlin mgraesslin at kde.org
Wed Jan 11 16:11:09 GMT 2017


Am 2017-01-11 13:40, schrieb Jan Kundrát:
> On středa 11. ledna 2017 6:57:50 CET, Martin Gräßlin wrote:
>> That doesn't work. Such inflexibility take away the advantage of 
>> having a CI.
> 
> What base system(s) do you prefer to target as a developer, Martin?

We release software which will be combined with a future set of 
software.
E.g. for KWin it has always been a problem that we develop against e.g.
Mesa X, but it will be combined by the distros with Mesa X+1, which 
hasn't
seen any testing from our side.

This resulted in lots of issues. For me as the maintainer of a high 
profile
application for our users the main need is to combine with the 
dependencies
the next version of KWin will use in real world.

This means: latest and greatest released software. Ideally fast rolling
release distributions.

Now that is for KWin master, for KWin stable or KWin LTS, the base 
doesn't
matter that much.

And I think KWin is there different to other software we have. We hardly
anywhere have the need for the latest and greatest. KWin needs it as it 
gets
integrated with it by the distros anyway.

An example is currently Xwayland. I would love to have a runtime 
dependency to the
latest released version as it fixes various issues. It will also be what 
will be used
by the next distributions, but we don't have it yet on our stable base. 
It means I have
to carry workarounds around, it means we have to put into the release 
announcement that
it should be combined with this version.

As I mentioned in another mail to this thread: KWin is pushing the 
stack.
Whether we like it or not. And I must say it's frustrating.

Cheers
Martin




More information about the kde-core-devel mailing list