Re: [platform-dev] Regressions in 4.12

[Aleksandar Kurtakov <akurtako@xxxxxxxxxx>]

On Wed, Mar 13, 2019 at 10:25 AM Sravan K Lakkimsetti <sravankumarl@xxxxxxxxxx> wrote:

Hi,

Let me clarify some details here

> (2) All test failures are "immediately" visible to a contributor/committer/... while contributing a change so he can improve the change before others see the failed tests. That seems to involve "waiting" for the SDK build right now and it is not only the contributor who sees the failures.

We have gerrit for this specific purpose. Gerrit builds do run tests. But the scope is limited to that git repository only. It does not run tests from other git repositories. Also the tests run on Linux only we don't have facility for running tests on Windows and Mac. One of the best practices we use is to get +1 from gerrit build before submitting.

Can you please raise a bug and propose you solution?

In this case we are speaking of missing junit test. There is ant.ui test that tests ant specific functionality which uncovers swt bug which  is not covered in swt test suite. Some of us spent multiple months per year (for multiple years now) babysitting the tests, reorganizing them, moving them to proper repositories so they are executed together with code changes not only nightly due to being in different repo, update tests to run with latest java versions, new dependencis and so on. It's a monster task which requires huge time investment and IMHO this is the least desired and appreciated work from contributors.

 

Thanks and Regards,
Sravan

Sravan Kumar Lakkimsetti
IBM India Pvt Ltd,
Embassy Golf Links Business Park, D Block,
Off Indiranagar-Kormangla Inner Ring Road,
Bangalore - 560071, India
Phone: 91-80-41776858

-----Original Message-----
From: Sebastian Lohmeier <sebastian@xxxxxxxxxxxxxxx>
Sent: Wednesday, March 13, 2019 12:59 PM
To: platform-dev@xxxxxxxxxxx
Subject: Re: [platform-dev] Regressions in 4.12

Hi,

I'm glad to see this thread!

Am 12.03.19 um 10:52 schrieb Andrey Loskutov:
[...]
> It would be great if the people contributing patches would also verify that the patches do not cause new regressions.
>
> Please after you have merged something (even if Gerrit is OK), consider to verify that your changes do not cause regressions on the next day SDK build.
[...]

(Unfortunately, my insights into the Eclipse build process is limited.) For me an ideal build infrastructure would have two properties:
(1) It makes test failures visible (that works already :-) ).
(2) All test failures are "immediately" visible to a contributor/committer/... while contributing a change so he can improve the change before others see the failed tests. That seems to involve "waiting" for the SDK build right now and it is not only the contributor who sees the failures.

Do others feel that (2) should/could be improved here? Does Bugzilla have issues for proposing solutions for (2) or even solving (2)? Is anyone working on this?

Thanks!

Sebastian

_______________________________________________
platform-dev mailing list
platform-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/platform-dev

_______________________________________________
platform-dev mailing list
platform-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/platform-dev

--

Alexander Kurtakov

Red Hat Eclipse Team