You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Mikhail Loenko <ml...@gmail.com> on 2007/03/22 07:14:15 UTC
[general] where we are with the "developers' releases"?
I have talked to some people who are interested in trying their apps on Harmony
and report back/fix the problems.
I'm reluctant to give them a link to the snapshots built from a random state.
What I suggest is: synchronize different Cruise Control runs.
So, some CC scenarios run more often, some run less often. Would it be
possible for the more often scenarios to include all thus revisions
where less often scenarios run plus something else?
In this case we can take the revisions where the least often scenario runs
and publish them as "tested snapshots" (if all the CCs were fine on
them of course)
I feel that it might be easily implemented: there might be a scheduler script
that monitors commits and put something on the Harmony web site
and CC infrastructure that monitors the website and run CCs accordingly.
Opinions?
Thanks,
Mikhail
Re: [general] where we are with the "developers' releases"?
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
On Mar 22, 2007, at 7:14 AM, Mikhail Loenko wrote:
> I have talked to some people who are interested in trying their
> apps on Harmony
> and report back/fix the problems.
>
> I'm reluctant to give them a link to the snapshots built from a
> random state.
Why would you think they are from a random state? The snaps should
pass the tests, as you have noted below.
>
> What I suggest is: synchronize different Cruise Control runs.
>
> So, some CC scenarios run more often, some run less often. Would it be
> possible for the more often scenarios to include all thus revisions
> where less often scenarios run plus something else?
We did talk about that - the less often would run on revisions that
passed the earlier stages, so you would know that we weren't testing
known crap.
>
> In this case we can take the revisions where the least often
> scenario runs
> and publish them as "tested snapshots" (if all the CCs were fine on
> them of course)
>
> I feel that it might be easily implemented: there might be a
> scheduler script
> that monitors commits and put something on the Harmony web site
> and CC infrastructure that monitors the website and run CCs
> accordingly.
>
> Opinions?
That's fine - this is good - but these still are "snapshots", not
"developers' releases"
geir
>
> Thanks,
> Mikhail
Re: [general] where we are with the "developers' releases"?
Posted by Mikhail Loenko <ml...@gmail.com>.
2007/3/22, Stepan Mishura <st...@gmail.com>:
> On 3/22/07, Mikhail Loenko wrote:
> > I have talked to some people who are interested in trying their apps on Harmony
> > and report back/fix the problems.
> >
> > I'm reluctant to give them a link to the snapshots built from a random state.
> >
>
> I guess you imply random = 'may be broken'. Actually, I monitor CC
> state and publish snapshots only if CC is up. Additionally often I run
you publish them manually???
> classlib tests (awt,swing excluded) for on snapshots.
>
> > What I suggest is: synchronize different Cruise Control runs.
> >
> > So, some CC scenarios run more often, some run less often. Would it be
> > possible for the more often scenarios to include all thus revisions
> > where less often scenarios run plus something else?
> >
>
> Not sure that I understood what you mean ... could you clarify?
well if short-cycle runs against revisions: 3, 5, 7, 12, 15, 19 and
long-cycle runs against 1, 11, 21, then it's not what I desire
if short-cycles run against 2, 4, 6, 8, 10, 12 and long cycle
against 4, 8, 12, then it's OK
>
> > In this case we can take the revisions where the least often scenario runs
> > and publish them as "tested snapshots" (if all the CCs were fine on
> > them of course)
> >
> > I feel that it might be easily implemented: there might be a scheduler script
> > that monitors commits and put something on the Harmony web site
> > and CC infrastructure that monitors the website and run CCs accordingly.
> >
> > Opinions?
> >
>
> From my point of view the problem is technical - synchronizing CC for
> different platforms. By 'synchronizing' I mean running CC on all
> platforms for the same svn revision. If all CC testing scenarios pass
> (long cycle) we publish snapshots for that revision as 'developer
> builds'. Does it works for you?
what I suggest is the script that mantains a number of lists:
list-1h will contain reivision numbers for CCs that run less than 1 hour,
list-2h ----//---- that run from 1 hour to 2 hour
list-1d for CCs run a whole day
CCs will take the revision numbers from that lists
Thanks,
Mikhail
>
> Thanks,
> Stepan Mishura
> Intel Enterprise Solutions Software Division
>
Re: [general] where we are with the "developers' releases"?
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
On Mar 22, 2007, at 5:13 AM, Mikhail Loenko wrote:
> 2007/3/22, Geir Magnusson Jr. <ge...@pobox.com>:
>>
>> On Mar 22, 2007, at 9:37 AM, Stepan Mishura wrote:
>>
>>
>> >>
>> >
>> > From my point of view the problem is technical - synchronizing
>> CC for
>> > different platforms. By 'synchronizing' I mean running CC on all
>> > platforms for the same svn revision. If all CC testing scenarios
>> pass
>> > (long cycle) we publish snapshots for that revision as 'developer
>> > builds'. Does it works for you?
>>
>> No. We can constrain our snapshots this way, but they are
>> snapshots. I'd like that our "developer builds" be real builds, with
>> targets of some sort, voting, etc.
>
> give it whatever name you want :)
> I'm talking about something up-to-date and stable
There's a big difference between a snapshot and a release. it's not
just me being silly about names...
geir
>
> Thanks,
> Mikhail
>
>>
>> geir
>>
>> >
>> > Thanks,
>> > Stepan Mishura
>> > Intel Enterprise Solutions Software Division
>>
>>
Re: [general] where we are with the "developers' releases"?
Posted by Mikhail Loenko <ml...@gmail.com>.
2007/3/22, Geir Magnusson Jr. <ge...@pobox.com>:
>
> On Mar 22, 2007, at 9:37 AM, Stepan Mishura wrote:
>
>
> >>
> >
> > From my point of view the problem is technical - synchronizing CC for
> > different platforms. By 'synchronizing' I mean running CC on all
> > platforms for the same svn revision. If all CC testing scenarios pass
> > (long cycle) we publish snapshots for that revision as 'developer
> > builds'. Does it works for you?
>
> No. We can constrain our snapshots this way, but they are
> snapshots. I'd like that our "developer builds" be real builds, with
> targets of some sort, voting, etc.
give it whatever name you want :)
I'm talking about something up-to-date and stable
Thanks,
Mikhail
>
> geir
>
> >
> > Thanks,
> > Stepan Mishura
> > Intel Enterprise Solutions Software Division
>
>
Re: [general] where we are with the "developers' releases"?
Posted by "Geir Magnusson Jr." <ge...@pobox.com>.
On Mar 22, 2007, at 9:37 AM, Stepan Mishura wrote:
>>
>
> From my point of view the problem is technical - synchronizing CC for
> different platforms. By 'synchronizing' I mean running CC on all
> platforms for the same svn revision. If all CC testing scenarios pass
> (long cycle) we publish snapshots for that revision as 'developer
> builds'. Does it works for you?
No. We can constrain our snapshots this way, but they are
snapshots. I'd like that our "developer builds" be real builds, with
targets of some sort, voting, etc.
geir
>
> Thanks,
> Stepan Mishura
> Intel Enterprise Solutions Software Division
Re: [general] where we are with the "developers' releases"?
Posted by Stepan Mishura <st...@gmail.com>.
On 3/22/07, Mikhail Loenko wrote:
> I have talked to some people who are interested in trying their apps on Harmony
> and report back/fix the problems.
>
> I'm reluctant to give them a link to the snapshots built from a random state.
>
I guess you imply random = 'may be broken'. Actually, I monitor CC
state and publish snapshots only if CC is up. Additionally often I run
classlib tests (awt,swing excluded) for on snapshots.
> What I suggest is: synchronize different Cruise Control runs.
>
> So, some CC scenarios run more often, some run less often. Would it be
> possible for the more often scenarios to include all thus revisions
> where less often scenarios run plus something else?
>
Not sure that I understood what you mean ... could you clarify?
> In this case we can take the revisions where the least often scenario runs
> and publish them as "tested snapshots" (if all the CCs were fine on
> them of course)
>
> I feel that it might be easily implemented: there might be a scheduler script
> that monitors commits and put something on the Harmony web site
> and CC infrastructure that monitors the website and run CCs accordingly.
>
> Opinions?
>
>From my point of view the problem is technical - synchronizing CC for
different platforms. By 'synchronizing' I mean running CC on all
platforms for the same svn revision. If all CC testing scenarios pass
(long cycle) we publish snapshots for that revision as 'developer
builds'. Does it works for you?
Thanks,
Stepan Mishura
Intel Enterprise Solutions Software Division