RFR: JDK-8170741: Enable uploading of built artifacts through Jib

Magnus Ihse Bursie magnus.ihse.bursie at oracle.com
Tue Dec 13 11:05:22 UTC 2016


On 2016-12-13 11:51, Erik Joelsson wrote:
> New webrev again: http://cr.openjdk.java.net/~erikj/8170741/webrev.03/
>
> Only a minor change to jprt.properties to enable testing of this 
> feature in JPRT.

Looks good to me.

/Magnus

>
> /Erik
>
>
> On 2016-12-09 15:55, Erik Joelsson wrote:
>> Hello
>>
>> New webrev: http://cr.openjdk.java.net/~erikj/8170741/webrev.02/
>>
>>
>> On 2016-12-08 14:19, Magnus Ihse Bursie wrote:
>>> On 2016-12-05 16:35, Erik Joelsson wrote:
>>>> For a while now, Oracle engineers have been able to configure 
>>>> builds of JDK 9 with binary dependencies automatically downloaded 
>>>> from an artifact storage, using Jib. Since then, Jib has been 
>>>> enhanced to also support publishing build artifacts produced by the 
>>>> build into the same storage. Such artifacts can subsequently be 
>>>> used as dependencies for other build targets.
>>>>
>>>> With this change, the Jib profiles configuration in JDK 9 are 
>>>> updated to leverage this new functionality in Jib. By doing this, 
>>>> we are able to express and define all deliverables from the build 
>>>> in the source code. We are also able to define possible workflows 
>>>> where one build step requires input from another.
>>>>
>>>> This change attempts to describe all artifacts currently built and 
>>>> stored by RE, with a few known exceptions that will be adjusted in 
>>>> followup fixes.
>>>>
>>>> With this change, there is a fair bit of refactoring in the 
>>>> jib-profiles.js file(s). The old designe resulted in a fair bit of 
>>>> duplicated configure arguments and it was hard to control what 
>>>> settings got duplicated in variants of profiles, such as *-debug. I 
>>>> believe the new patterns are more flexible without sacrificing 
>>>> readability. I have manually inspected the generated configure 
>>>> lines for all relevant profiles and compared before and after the 
>>>> change to make sure no regressions are introduced by this.
>>>>
>>>> I moved more of the default version numbers into the 
>>>> version-numbers file so that they can be read from jib-profiles.js.
>>>>
>>>> In langtools/test/Makefile, I made it possible to override the 
>>>> TEST_OUTPUT_DIR, like it is currently possible for the other test 
>>>> makefiles.
>>>>
>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8170741
>>>>
>>>> Webrev: http://cr.openjdk.java.net/~erikj/8170741/webrev.01/
>>>
>>> Looks good to me overall, but I do not like the massive amount of 
>>> code duplication in profilesArtifacts. These could, and should, be 
>>> generated from some suitable input (like name of os and cpu in the 
>>> resulting bundle, etc)
>>>
>> Fixed.
>>> The new print-config.js tool, is it used somewhere, or is it just 
>>> provided as a development tool for adhoc use? Is common/conf really 
>>> the proper place for it, and not common/bin?
>>>
>> Good point, moved.
>>
>> /Erik
>




More information about the build-dev mailing list