harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stepan Mishura" <stepan.mish...@gmail.com>
Subject Re: [build?] Problem with snapshot build or download page?
Date Fri, 19 Oct 2007 06:23:26 GMT
On 10/18/07, Sian January <sianjanuary@googlemail.com> wrote:
> Hi Stepan,
>
> Thanks for your reply.  Yes - I was meaning successful as not only compiled
> succesfully but passed all the required tests.  Required tests would be a
> set of tests that we aimed to be GREEN all the time.  Then if something is
> checked in that breaks one of these tests it would either be fixed fairly
> quickly (say within a day or two?) or rolled back.  We could publish any (or
> every) build that passed this subset, which should mean that there's never a
> long period of time without a build being published.
>
> I think this would also help with reducing the freeze period for a milestone
> because it would mean we would never get too far away from something that's
> working, so stabilization would be easier.  It's also easier to fix
> something that's just been committed than something that you did 2 or 3
> months ago.
>
> I like the suggestion of using the integrity suite as a start - it seems
> like a good basic set of tests to aim for. But if you don't think it's
> realistic to keep these green most the time during a development cycle, then
> maybe just start with the classlib, drlvm and drlvm-regression tests?  These
> are all green at the moment on 32-bit Linux and Windows and I think the
> Harmony community is already quite good at trying to keep those tests
> passing.
>

I agree with 32-bit Linux and Windows platforms. But I think that a
set of suites should include Eclipse Hello World Application (EHWA)
suite and may be jdktools too. Say if EHWA is broken than most
probably that there is serious regression. And IMHO such build
shouldn't be published as users build - it can not run simple Eclipse
scenario.

Thanks,
Stepan.

> Thanks,
>
> Sian
>
>
>
> On 18/10/2007, Stepan Mishura <stepan.mishura@gmail.com> wrote:
> >
> > On 10/17/07, Sian January <sianjanuary@googlemail.com> wrote:
> > > That's good - it sounds like we are aiming for the same thing after
> > all!  It
> > > would be nice if this could be automated at some point, so successful
> > builds
> > > are published automatically.  Also should the build be considered broken
> > if
> > > we have an unsuccessful build?  It just seems to me like something is
> > wrong
> > > if there hasn't been a 'successful' build for more than 2 weeks.  What
> > do
> > > you think?
> > >
> >
> > Hi Sian,
> >
> > Do you mean 'successful'=='tested' build?
> >
> > I agree that we need to work out a criteria for publishing tested
> > builds for users. Of cause if everything became GREEN on the testing
> > page then the build is published as users build. But it is not so
> > probable to see during active code development. Also I agree that we
> > should define time frame - if we can not publish users build for a
> > long time that we need to stop and fix all failures. I'd suggest to
> > use a set of suites run by integrity testing [1]. So if classlib
> > tests, drlvm and drlvm-regression tests, jdktools and Eclipse hello
> > world application pass on snapshot testing them we publish the build.
> >
> > [1] http://people.apache.org/~varlax/harmony-integrity/
> >
> > Thanks,
> > Stepan.
> >
> > <SNIP>
> >
>
>
>
> --
> Unless stated otherwise above:
> IBM United Kingdom Limited - Registered in England and Wales with number
> 741598.
> Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU

Mime
View raw message