How to run tests before pushing a backport changeset?

[Andrew Haley]

On 2/21/19 9:07 AM, Aleksey Shipilev wrote:
> On 2/21/19 4:55 AM, Martin Buchholz wrote:
>> On Wed, Feb 20, 2019 at 4:10 PM Aleksey Shipilev <shade at redhat.com <mailto:shade at redhat.com>> wrote:

>>     > I read the links on
>>     > https://openjdk.java.net/projects/jdk-updates/, but couldn't
>>     > find any description on testing. Does the backporting process
>>     > require running such automated tests before pushing a
>>     > changeset?

>>     I think it is maintainer's duty to decide if testing was enough
>>     (this is why Fix Request requires [1] to spell out what testing
>>     was done). That said, it seems the more tests you run, the more
>>     less chance backport would be problematic.

>> I'm surprised to see you say that, because you are the king of
>> build/test automation.  I prefer to lessen the burden on backport
>> developers and instead invest in better release testing automation
>> (e.g. "presubmit queue").
> 
> By all means, you are welcome to invest in better release test
> automation :)
>
> OpenJDK historically relies heavily on developers doing
> pre-integration testing themselves, and the bulk of tests are
> running in post-integration time. JDK Updates project is not the
> exception to this. Maybe Skara would make it better, but this
> remains to be seen.
> 
> There are many tests and test suites to choose from, "tier1" seems
> to be universally used as the go-to pre-integration test suite. If
> tier1 fails, it is likely every other developer would complain about
> bugs once you push. If tier1 passes, it is likely there are no bugs
> on frequent product paths.

Yes, and we'd like to work with AdoptOpenJDK to integrate their testing
into the OpenJDK workflow. It's a work in progress.

-- 
Andrew Haley
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671