Re: Build windows-jdk15

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

Elliotte Rusty Harold
On Wed, Dec 30, 2020 at 8:58 PM John Patrick <[hidden email]> wrote:
>
> Better to know the issues and slowly fix them then get hit with a load
> of issues blocking users.
>

We have the Linux JDK 15, and Windows JDK 16, builds for that. The
windows JDK-15 build is noise that provides no useful information.

--
Elliotte Rusty Harold
[hidden email]

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

rfscholte
What's you analysis?
Did you inform INFRA?
If we have issues with these versions, most likely Maven users will have to.

We've seen this in the past too and I think all the time I'm been able to bring the builds back to blue.
Disabling would be avoiding or ignoring the problem.

thanks,
Robert
On 30-12-2020 14:59:16, Elliotte Rusty Harold <[hidden email]> wrote:
The Windows JDK 15 instance in jenkins seems irredeemably flaky. It
fails more often than not for reasons that have nothing to do with the
code being tested.

As JDK 15 is not a long term support release, and has only minimal
adoption, can we simply turn off the Windows JDK 15 build?

Personally, I'd love to turn off all JDK 15 and 16 builds, but Windows
JDK 15 is by far the worst.

--
Elliotte Rusty Harold
[hidden email]

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

rfscholte
In reply to this post by Elliotte Rusty Harold
Great, thanks!

I expect that one of the other Jenkins jobs overwrites JAVA_HOME, not knowing it effects all following jobs.

Robert
On 31-12-2020 09:24:07, Benjamin Marwell <[hidden email]> wrote:
Here you go: https://issues.apache.org/jira/browse/INFRA-21244

Already opened a week ago when Hervé tried to merge a PR and suddenly the
build failed.

Disabling JDK15 might be a valid solution if they don't react soon.



On Wed, 30 Dec 2020, 14:59 Elliotte Rusty Harold,
wrote:

> The Windows JDK 15 instance in jenkins seems irredeemably flaky. It
> fails more often than not for reasons that have nothing to do with the
> code being tested.
>
> As JDK 15 is not a long term support release, and has only minimal
> adoption, can we simply turn off the Windows JDK 15 build?
>
> Personally, I'd love to turn off all JDK 15 and 16 builds, but Windows
> JDK 15 is by far the worst.
>
> --
> Elliotte Rusty Harold
> [hidden email]
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>
>
Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

Elliotte Rusty Harold
In reply to this post by rfscholte
On Wed, Dec 30, 2020 at 11:38 PM Robert Scholte <[hidden email]> wrote:
>
> What's you analysis?
> Did you inform INFRA?
> If we have issues with these versions, most likely Maven users will have to.
>

The issues all seem to be in the test infrastructure, and have nothing
to do with the Maven code itself. While it's conceivable for a build
to find a bug that only surfaces on one platform/JDK combination, this
is not what we see in practice. Most bugs trigger in a particular JDK
XOR on a particular platform. I'm not sure I've ever noticed a genuine
Maven bug that triggered in only one build.

Diagnosing these is extremely challenging given the Jenkins UI. It is
very hard to find the logs for any particular job. They are very
hidden.

Various people have reported and fixed these INFRA issues before. It
doesn't stick. I prefer fewer, more reliable CI builds.

--
Elliotte Rusty Harold
[hidden email]

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

rfscholte
In reply to this post by rfscholte
Have you tried https://ci-builds.apache.org/blue/organizations/jenkins/Maven%2Fmaven-box%2Fmaven-dependency-plugin/detail/null/2/pipeline ?
It shows the output of the failing IT. It expects verbose information, but it is missing.


Robert
On 31-12-2020 13:29:54, Elliotte Rusty Harold <[hidden email]> wrote:
I'm not sure we're looking at the same thing.
https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/
shows three runs of the same most recently committed code. The first
two failed flakily on different instances and passed on all other
instances. The third passed on all instances. Master is also passing
and has been for the last few days/commits.Where are you seeing a
failure?

On Thu, Dec 31, 2020 at 12:13 PM Robert Scholte wrote:

>
> Looks like the IT for is failing MDEP-644[1].
> Elliotte, you closed this one, but I'm missing some important information in this ticket: a link to the commit.
> Now it is hard to figure out if there's a SNAPSHOT issue in play.
> You see these kind of failures if branches of a SNAPSHOT dependency use the same version.
> This would be my first guess.
> But the failure is correct and is most likely unrelated to INFRA, but an issue with either the POM or the IT.
>
> Robert
>
>
> [1] https://issues.apache.org/jira/browse/MDEP-644
> On 31-12-2020 12:48:26, Bernd Eckenfels wrote:
> Hello,
>
> As long as we have windows builds and jdk15+ builds I guess one can reduce the flaky combinations but it would still be interesting to see what’s the problem is, in the pipeline steps view you see the failed step
>
> https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/2/flowGraphTable/
>
> And it’s console log
>
> https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/2/execution/node/315/log/
>
> I complains about a unexpected dependency tree (this does not really look platform-version specific, maybe more a repository issue?)
>
> Gruss
> Bernd
> --
> http://bernd.eckenfels.net
> ________________________________
> Von: Elliotte Rusty Harold
> Gesendet: Thursday, December 31, 2020 12:37:18 PM
> An: Maven Developers List
> Betreff: Re: Build windows-jdk15
>
> Here's one example:
>
> https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/
>
> There are many others. Nothing changed between the three runs. First
> Windows JDK 15 failed. Then Windows JDK 7 failed. Then they all
> passed. No code changes were made between the runs. If there's a way
> to see the logs from the failed runs, it's far from obvious to me.
>
> Flaky build infrastructure is dangerous because it teaches developers
> to ignore the CI.
>
>
>
>
> On Wed, Dec 30, 2020 at 11:38 PM Robert Scholte wrote:
> >
> > What's you analysis?
> > Did you inform INFRA?
> > If we have issues with these versions, most likely Maven users will have to.
> >
> > We've seen this in the past too and I think all the time I'm been able to bring the builds back to blue.
> > Disabling would be avoiding or ignoring the problem.
> >
> > thanks,
> > Robert
> > On 30-12-2020 14:59:16, Elliotte Rusty Harold wrote:
> > The Windows JDK 15 instance in jenkins seems irredeemably flaky. It
> > fails more often than not for reasons that have nothing to do with the
> > code being tested.
> >
> > As JDK 15 is not a long term support release, and has only minimal
> > adoption, can we simply turn off the Windows JDK 15 build?
> >
> > Personally, I'd love to turn off all JDK 15 and 16 builds, but Windows
> > JDK 15 is by far the worst.
> >
> > --
> > Elliotte Rusty Harold
> > [hidden email]
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [hidden email]
> > For additional commands, e-mail: [hidden email]
> >
>
>
> --
> Elliotte Rusty Harold
> [hidden email]
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>


--
Elliotte Rusty Harold
[hidden email]

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Build windows-jdk15

Elliotte Rusty Harold
And the next run everything passed without any code changes:

https://ci-builds.apache.org/blue/organizations/jenkins/Maven%2Fmaven-box%2Fmaven-dependency-plugin/detail/null/3/pipeline




On Thu, Dec 31, 2020 at 6:04 PM Robert Scholte <[hidden email]> wrote:

>
> Have you tried https://ci-builds.apache.org/blue/organizations/jenkins/Maven%2Fmaven-box%2Fmaven-dependency-plugin/detail/null/2/pipeline ?
> It shows the output of the failing IT. It expects verbose information, but it is missing.
>
>
> Robert
> On 31-12-2020 13:29:54, Elliotte Rusty Harold <[hidden email]> wrote:
> I'm not sure we're looking at the same thing.
> https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/
> shows three runs of the same most recently committed code. The first
> two failed flakily on different instances and passed on all other
> instances. The third passed on all instances. Master is also passing
> and has been for the last few days/commits.Where are you seeing a
> failure?
>
> On Thu, Dec 31, 2020 at 12:13 PM Robert Scholte wrote:
> >
> > Looks like the IT for is failing MDEP-644[1].
> > Elliotte, you closed this one, but I'm missing some important information in this ticket: a link to the commit.
> > Now it is hard to figure out if there's a SNAPSHOT issue in play.
> > You see these kind of failures if branches of a SNAPSHOT dependency use the same version.
> > This would be my first guess.
> > But the failure is correct and is most likely unrelated to INFRA, but an issue with either the POM or the IT.
> >
> > Robert
> >
> >
> > [1] https://issues.apache.org/jira/browse/MDEP-644
> > On 31-12-2020 12:48:26, Bernd Eckenfels wrote:
> > Hello,
> >
> > As long as we have windows builds and jdk15+ builds I guess one can reduce the flaky combinations but it would still be interesting to see what’s the problem is, in the pipeline steps view you see the failed step
> >
> > https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/2/flowGraphTable/
> >
> > And it’s console log
> >
> > https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/2/execution/node/315/log/
> >
> > I complains about a unexpected dependency tree (this does not really look platform-version specific, maybe more a repository issue?)
> >
> > Gruss
> > Bernd
> > --
> > http://bernd.eckenfels.net
> > ________________________________
> > Von: Elliotte Rusty Harold
> > Gesendet: Thursday, December 31, 2020 12:37:18 PM
> > An: Maven Developers List
> > Betreff: Re: Build windows-jdk15
> >
> > Here's one example:
> >
> > https://ci-builds.apache.org/job/Maven/job/maven-box/job/maven-dependency-plugin/job/null/
> >
> > There are many others. Nothing changed between the three runs. First
> > Windows JDK 15 failed. Then Windows JDK 7 failed. Then they all
> > passed. No code changes were made between the runs. If there's a way
> > to see the logs from the failed runs, it's far from obvious to me.
> >
> > Flaky build infrastructure is dangerous because it teaches developers
> > to ignore the CI.
> >
> >
> >
> >
> > On Wed, Dec 30, 2020 at 11:38 PM Robert Scholte wrote:
> > >
> > > What's you analysis?
> > > Did you inform INFRA?
> > > If we have issues with these versions, most likely Maven users will have to.
> > >
> > > We've seen this in the past too and I think all the time I'm been able to bring the builds back to blue.
> > > Disabling would be avoiding or ignoring the problem.
> > >
> > > thanks,
> > > Robert
> > > On 30-12-2020 14:59:16, Elliotte Rusty Harold wrote:
> > > The Windows JDK 15 instance in jenkins seems irredeemably flaky. It
> > > fails more often than not for reasons that have nothing to do with the
> > > code being tested.
> > >
> > > As JDK 15 is not a long term support release, and has only minimal
> > > adoption, can we simply turn off the Windows JDK 15 build?
> > >
> > > Personally, I'd love to turn off all JDK 15 and 16 builds, but Windows
> > > JDK 15 is by far the worst.
> > >
> > > --
> > > Elliotte Rusty Harold
> > > [hidden email]
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: [hidden email]
> > > For additional commands, e-mail: [hidden email]
> > >
> >
> >
> > --
> > Elliotte Rusty Harold
> > [hidden email]
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [hidden email]
> > For additional commands, e-mail: [hidden email]
> >
>
>
> --
> Elliotte Rusty Harold
> [hidden email]
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>


--
Elliotte Rusty Harold
[hidden email]

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]