Re: Possibility to increase release frequency - Mailing list pgadmin-hackers

On Mon, Apr 30, 2018 at 10:51 AM, Dave Page <dpage@pgadmin.org> wrote:
> Hi
>
> On Mon, Apr 30, 2018 at 3:33 PM, Robert Eckhardt <reckhardt@pivotal.io>
> wrote:
>>
>> All,
>>
>> Currently we are starting to get a fair number of users leveraging
>> pgAdmin 4. Because of this we are finding new issues with Greenplum we
>> didn't previously know about and we would like to get the fixes for
>> those issues out as soon as possible.
>>
>> The current release process is shrouded in a bit of darkness for us so
>> I'm not sure what this ask even entails.
>
>
> I'm actively working on a buildfarm for doing automated official releases
> that folks other than I will have access to. Current status:
>
> - Jenkins host up and running
> - Automated dependency builds of zlib, OpenSSL and PostgreSQL running on
> Windows (just pgAdmin to go)
> - Automated dependency builds of OpenSSL and PostgreSQL running on macOS
> (just pgAdmin to go)
> - Automated dependency builds of PostgreSQL running on Linux
> - All test suites (JS linter and tests, Python PEP-8, unit/API and feature
> tests), components (message catalogs, docs, Qt4 & Qt5 runtime builds) and
> builds of pgAdmin (source, Python wheel) running on Linux and being tested
> against PG 9.3 - 10 and EPAS 9.4 - 10. I have had a (singular) successful
> feature test run, but most of the time they just time out at the moment.
>
> Once I have all the builds working as required for each platform, I intend
> to have them run regularly, and then setup smaller, targeted builds that
> will just produce the distribution packages on demand.
>
>>
>> Ask: what can we do to accelerate the release process? Can we help
>> automate builds or reduce the QA load or anything like that?
>>
>> Ideally we'd like to be releasing weekly, how can we move in that
>> direction?
>
>
> Right now we're still relying far too much on Fahar's manual testing for
> that to be a reality. He covers a lot of platform-specific tests on
> different distros with every release that often pickup showstoppers.

Makes perfect sense.

>
> I think the best way forward would be to continue with the Electron work so
> we can get away from using the browser in Desktop mode, and then once we
> have a stable and consistent desktop mode, continue to work on improving
> test coverage to minimise reliance on Fahar.

Electron is the next thing to get picked up by us.

Can you provide any insight into what Fahar does, one of the things
I've discussed is actually testing the installers after they have been
built. We didn't really have a plan for that so much so looking at
what was currently done could help generate some ideas.