Do you guys want to try this on something in commons first? We could try it on 
a lesser used component to see how it performs.

-Rob

> On Mar 31, 2019, at 8:25 PM, Matt Sicker <[email protected]> wrote:
> 
> It files a PR which kicks off a CI build. I think you can configure it to
> automatically merge if the build passes. Otherwise it’s jus another open PR.
> 
> On Sun, Mar 31, 2019 at 17:17, Gary Gregory <[email protected]> wrote:
> 
>> Might be worth a try.  I take it it only flies PRs if a build passes all
>> tests?
>> 
>> Gary
>> 
>> On Mon, Mar 25, 2019, 14:23 Matt Sicker <[email protected]> wrote:
>> 
>>> I wasn't thinking of automating this fully at any point. The bot does
>>> file PRs for each dependency individually. I'm not sure how
>>> configurable everything is, but I do like the sound of having multiple
>>> dependency management profiles for compatibility testing purposes.
>>> 
>>> On Mon, 25 Mar 2019 at 13:07, Gary Gregory <[email protected]>
>> wrote:
>>>> 
>>>> This reminds me: What happened to Apache Gump? Didn't it handle this?
>>>> 
>>>> Gary
>>>> 
>>>> On Mon, Mar 25, 2019 at 2:02 PM Gary Gregory <[email protected]>
>>> wrote:
>>>> 
>>>>> There room for automation in this area, but I only seen it as an
>>>>> informative step for now. Like the tool saying "I've ran the build
>> with
>>>>> this change and it's green/red". Today, without this tool, I could
>> see
>>> a
>>>>> tool doing :
>>>>> - create a branch for a change and push it, create a PR
>>>>> - look at what GitHub/Travis-CI reports
>>>>> 
>>>>> then you, the dev:
>>>>> 
>>>>> - accept/reject the PR
>>>>> 
>>>>> This would only be done one dep change at a time. IOW, only once
>> you've
>>>>> accepted/rejected a dep change would you move on to the next one.
>>>>> 
>>>>> Gary
>>>>> 
>>>>> On Mon, Mar 25, 2019 at 1:46 PM Ralph Goers <
>>> [email protected]>
>>>>> wrote:
>>>>> 
>>>>>> I have mixed feelings. Gary has been manually doing this for quite a
>>>>>> while and I have always had concerns - besides driving his commit
>>> count
>>>>>> artificially through the roof.
>>>>>> 
>>>>>> Managing dependencies really needs to be managed carefully. When we
>>>>>> upgrade a dependency that effectively becomes the minimum supported
>>>>>> version. In that sense we are upgrading dependencies way too fast.
>> At
>>> the
>>>>>> same time, we need to insure that we don’t have problems with new
>>> releases.
>>>>>> As such we really need two versions of the parent pom - one that is
>>> used
>>>>>> for verifying the latest versions of dependencies and one that is
>>> used for
>>>>>> the release that specifies the minimum supported version (typically
>>> this
>>>>>> would be expressed as a version range).
>>>>>> 
>>>>>> I’m also concerned because some dependencies upgrade their minimum
>>>>>> required JDK version in a minor release. That happened with the
>> Flume
>>> 1.8
>>>>>> release. So we cannot upgrade to that version in the release-2.x
>>> branch,
>>>>>> although our users can if they want to.  We also ran into a problem
>>> with
>>>>>> SLF4J. The latest release dropped a class that we use. We have
>>> modified the
>>>>>> code to support the latest releases but we require the last release
>>> that
>>>>>> had the class to compile.
>>>>>> 
>>>>>> Also, our process has always been to create a Jira for everything,
>>>>>> including updating dependency versions, and including them in
>>> changes.xml.
>>>>>> It looks like this tool doesn’t do either of these things.
>>>>>> 
>>>>>> Ralph
>>>>>> 
>>>>>>> On Mar 25, 2019, at 9:24 AM, Matt Sicker <[email protected]>
>> wrote:
>>>>>>> 
>>>>>>> Hi all,
>>>>>>> 
>>>>>>> Various Jenkins projects have been using Dependabot [1] to
>>>>>>> automatically make PRs to update dependencies. We could use this
>> for
>>>>>>> most of our components it looks like. What do you think about
>>> adopting
>>>>>>> use of this bot?
>>>>>>> 
>>>>>>> [1]: https://dependabot.com/
>>>>>>> 
>>>>>>> --
>>>>>>> Matt Sicker <[email protected]>
>>>>>>> 
>>>>>> 
>>>>>> 
>>> 
>>> 
>>> 
>>> --
>>> Matt Sicker <[email protected]>
>>> 
>> 
> -- 
> Matt Sicker <[email protected]>

Reply via email to