You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by Mariia Mykhailova <ma...@microsoft.com> on 2015/10/30 23:38:45 UTC

Jenkins builds failing to set build status on GitHub commit

Hi,

I've noticed that we get a lot of build failures in https://builds.apache.org/job/Reef-pull-request-ubuntu/ . In all of them build actually succeeds and tests pass, but Jenkins fails to set "success" build status on pull request, so it sets failed status instead.

Here is an example: https://builds.apache.org/job/Reef-pull-request-ubuntu/789/


Setting commit status on GitHub for https://github.com/apache/incubator-reef/commit/be0995a7f7e3448d7fa14a4a4f427a59d2fd364a

ERROR: Publisher 'Set build status on GitHub commit' aborted due to exception:

java.io.IOException<http://stacktrace.jenkins-ci.org/search?query=java.io.IOException>: {"message":"No commit found for SHA: be0995a7f7e3448d7fa14a4a4f427a59d2fd364a"

Jenkins is looking for commit be0995a7f7e3448d7fa14a4a4f427a59d2fd364a, but the real commit is dfe86d5204b523ef5d85177aa677b4a2fe7b4548 (https://github.com/apache/incubator-reef/commit/dfe86d5204b523ef5d85177aa677b4a2fe7b4548). Successful builds have these hashes matching.

What could be the cause of hash mismatch? Does Git use source code as committed, while Jenkins merges it with latest master, so there is difference if pull request is published without rebasing it to master it first? It would be nice if we could fix the cause, so that our Jenkins builds fail only if there is a real issue.

-Mariia

Re: Jenkins builds failing to set build status on GitHub commit

Posted by Dongjoon Hyun <do...@apache.org>.
That sounds great!

Dongjoon.

On Tue, Nov 3, 2015 at 4:37 AM, Mariia Mykhailova <ma...@microsoft.com>
wrote:

> This might be one cause, but not the only one.
> The build I sent as an example,
> https://builds.apache.org/job/Reef-pull-request-ubuntu/789/
> corresponds to a pull request which was not rebased by us but still
> failed, https://github.com/apache/incubator-reef/pull/603
>
> In an offline talk Marcus suggested that we disable Jenkins for Ubuntu,
> since we have a more stable Travis build to cover this.
>
> -Mariia
>
> -----Original Message-----
> From: Markus Weimer [mailto:markus@weimo.de]
> Sent: Sunday, November 1, 2015 6:59 AM
> To: dev@reef.incubator.apache.org
> Subject: Re: Jenkins builds failing to set build status on GitHub commit
>
> On 2015-10-30 15:38, Mariia Mykhailova wrote:
> > Jenkins is looking for commit
> > be0995a7f7e3448d7fa14a4a4f427a59d2fd364a, but the real commit is
> > dfe86d5204b523ef5d85177aa677b4a2fe7b4548
> > (
> https://na01.safelinks.protection.outlook.com/?url=https%3a%2f%2fgithub.com%2fapache%2fincubator-reef%2fcommit%2fdfe86d5204b523ef5d85177aa677b4a2fe7b4548&data=01%7c01%7cmamykhai%40microsoft.com%7c09eb55a14301443d727908d2e2cd4b7a%7c72f988bf86f141af91ab2d7cd011db47%7c1&sdata=NEerx1CSHT%2f%2bf%2fgUwlLVHxppPNikDaWpuEwh9FUT7IM%3d
> ).
> > Successful builds have these hashes matching.
> >
> > What could be the cause of hash mismatch? Does Git use source code as
> > committed, while Jenkins merges it with latest master, so there is
> > difference if pull request is published without rebasing it to master
> > it first? It would be nice if we could fix the cause, so that our
> > Jenkins builds fail only if there is a real issue.
>
> My hunch is that this happens whenever we rebase a PR after it has been
> opened. That indeed changes the hash.
>
> Markus
>

RE: Jenkins builds failing to set build status on GitHub commit

Posted by Mariia Mykhailova <ma...@microsoft.com>.
This might be one cause, but not the only one.
The build I sent as an example, https://builds.apache.org/job/Reef-pull-request-ubuntu/789/
corresponds to a pull request which was not rebased by us but still failed, https://github.com/apache/incubator-reef/pull/603

In an offline talk Marcus suggested that we disable Jenkins for Ubuntu, since we have a more stable Travis build to cover this.

-Mariia

-----Original Message-----
From: Markus Weimer [mailto:markus@weimo.de] 
Sent: Sunday, November 1, 2015 6:59 AM
To: dev@reef.incubator.apache.org
Subject: Re: Jenkins builds failing to set build status on GitHub commit

On 2015-10-30 15:38, Mariia Mykhailova wrote:
> Jenkins is looking for commit
> be0995a7f7e3448d7fa14a4a4f427a59d2fd364a, but the real commit is
> dfe86d5204b523ef5d85177aa677b4a2fe7b4548
> (https://na01.safelinks.protection.outlook.com/?url=https%3a%2f%2fgithub.com%2fapache%2fincubator-reef%2fcommit%2fdfe86d5204b523ef5d85177aa677b4a2fe7b4548&data=01%7c01%7cmamykhai%40microsoft.com%7c09eb55a14301443d727908d2e2cd4b7a%7c72f988bf86f141af91ab2d7cd011db47%7c1&sdata=NEerx1CSHT%2f%2bf%2fgUwlLVHxppPNikDaWpuEwh9FUT7IM%3d).
> Successful builds have these hashes matching.
> 
> What could be the cause of hash mismatch? Does Git use source code as 
> committed, while Jenkins merges it with latest master, so there is 
> difference if pull request is published without rebasing it to master 
> it first? It would be nice if we could fix the cause, so that our 
> Jenkins builds fail only if there is a real issue.

My hunch is that this happens whenever we rebase a PR after it has been opened. That indeed changes the hash.

Markus

Re: Jenkins builds failing to set build status on GitHub commit

Posted by Markus Weimer <ma...@weimo.de>.
On 2015-10-30 15:38, Mariia Mykhailova wrote:
> Jenkins is looking for commit
> be0995a7f7e3448d7fa14a4a4f427a59d2fd364a, but the real commit is
> dfe86d5204b523ef5d85177aa677b4a2fe7b4548
> (https://github.com/apache/incubator-reef/commit/dfe86d5204b523ef5d85177aa677b4a2fe7b4548).
> Successful builds have these hashes matching.
> 
> What could be the cause of hash mismatch? Does Git use source code as
> committed, while Jenkins merges it with latest master, so there is
> difference if pull request is published without rebasing it to master
> it first? It would be nice if we could fix the cause, so that our
> Jenkins builds fail only if there is a real issue.

My hunch is that this happens whenever we rebase a PR after it has been
opened. That indeed changes the hash.

Markus