[Vm-dev] Status of VM

[Ben Coman]

Hi Nicolas,

On 3 April 2018 at 13:42, Nicolas Cellier
<nicolas.cellier.aka.nice at gmail.com> wrote:
>
> Hi all,
>
> I'm not mandated by opensmalltalk organization nor anyone but myself. But as a member I want to give my feelings.

It can be tough to do so. I'm sure most here appreciate your effort to do so.


> We also must take care of the infrastructure, and our first duty is to not let the CI rot in red, otherwise there's no process possible at all.

Its easy to get stuck in "what really happens with CI"
http://www.yegor256.com/2014/10/08/continuous-integration-is-dead.html#what-happens-in-reality

Apart from helping to detect new errors quickly, keeping the build
green has an encouraging effect on occasional VM'ers
and attracting new contributors.


hey, cool !  Its actually green right now...
https://travis-ci.org/OpenSmalltalk/opensmalltalk-vm/builds/361738726

btw, thanks Fabio for setting up the stages. Apart from other
technical reasons, the grouped presentation looks real nice.


> I wanted to have a look at this status which is red again. It's not catastrophic, it was caused by a gcc bug and affected the pharo.sista.spur and pharo.cog.spur.lowcode brands. I'm very happy to see all the goodwill with tips and encouragement I received in a single day from Holger, Alistair, Ben, Subbu, Fabio, Tobias, with their help, that's solved. But it's not finished.
>
> The next step is to solve the deploy step that just failed for Pharo (https://travis-ci.org/OpenSmalltalk/opensmalltalk-vm/jobs/361346561) and also the bintray problems which I do not understand and concern everyone (https://ci.appveyor.com/project/OpenSmalltalk/vm/build/1.0.1219/job/2j1bu778vg9a2rpw). More help wanted...

Perhaps it would be useful to separate deployment into a stage of its
own, following immediately after "Main squeak and pharo builds" stage.
The deployment jobs might be "Allow to fail", to limit the impact on
the broader community.
Its the "building" / "testing" that is important to keep green across
the community.

If the deployment stage is conditional on *all* of the Main jobs
succeeding (i.e. the default)
it shouldn't slow things down too much due to the smaller Main stage,
and might encourage breakage to be fixed/reverted asap.

cheers -ben