flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christofer Dutz <christofer.d...@c-ware.de>
Subject AW: [FLEXJS][FALCONJX] Builds and CI (was Re: [DISCUSS] Release Apache FlexJS 0.5.0
Date Tue, 06 Oct 2015 06:58:58 GMT
Well I think that's exactly what jenkins is built for ... all these sequences of things that
are too boring to execute every time. 
You could have one "freestyle" job with mutliple steps and the last one being "deploy" after
step 5 (in justins list) has passed. Or you have multiple jobs ... for example flex-falcon
and flex-sdk which each have the "mustella" test run against them automatically.

In my current project here we have a network of about 70 independent Jenkins jobs that each
trigger another in sequence ... but they handle everything from building, unti-testing, integration-testing,
acceptence-testing, deployment, ... all working niceley.


Von: Alex Harui <aharui@adobe.com>
Gesendet: Dienstag, 6. Oktober 2015 06:21
An: dev@flex.apache.org
Betreff: [FLEXJS][FALCONJX] Builds and CI (was Re: [DISCUSS] Release Apache FlexJS 0.5.0

On 10/5/15, 5:19 PM, "Justin Mclean" <justin@classsoftware.com> wrote:

>> I think you may not have picked up the full context of the question.  If
>> you have a repo of framework code, its automated tests need the compiler
>> to compile them.  Meanwhile repo of compiler code has an automated test
>> prove that any change to the compiler can still compile the framework.
>Sounds like you just need to split the compiler tests into two sets and
>run in this order:
>1. compile the compiler
>2. run the non framework compiler tests
>3. use the compiler to compile the framework
>4. run the framework compiler tests
>5. run the framework tests

Yes, that is the correct sequence when starting from scratch or when both
repos have interdependent changes.  But for most changes, we don’t need
five distinct steps.  Should we have 5 jenkins jobs?  Actually, we’d need
two more to produce nighty builds once step 5 is successful.  Otherwise a
bad compiler nightly could get consumed by someone.

What do other’s think?  No objection from me if someone has the time to do
it.  I think we have all the right ant targets so it is just a matter of
setting up a different set of jobs.


View raw message