You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@maven.apache.org by "Rob @ DNR" <rv...@dotnetrdf.org> on 2025/12/09 10:32:05 UTC
CI Skip Breaking Change in maven-release-plugin 3.3.0
Hey Maven folks,
I wanted to raise concerns about a breaking change made in the 3.3.0 release of the maven-release-plugin
This change was suggested in https://github.com/apache/maven-release/issues/1422 and provided in PR https://github.com/apache/maven-release/pull/1423
Despite several well-founded objections to this idea being made in the original issue, and the PR, these were dismissed by the OP (a Maven PMC member AFAICT) and this change was merged with seemingly minimal discussion or consideration of its impacts on the many developers who depend on Maven for their builds.
If you weren’t already aware this breaking change adds the "[ci skip]” text snippet into the commit message that the release plugin makes for the release commit, i.e. the SCM commit that has the non-SNAPSHOT version set and is tagged with the release tag. The effect of this text is that most common CI/CD systems ignore this commit and don’t build it.
However, many developers and organisations, myself included, have designed their CI/CD workflows around triggering release workflows from the release commit which this change breaks. Forcing us either to change our plugin configuration back to the old commit message default, or rollback to an earlier plugin version.
Why was a breaking change of this magnitude not raised on the developers list and discussed more widely?
This is in fact the second maven-release-plugin release in a row that has been unusable for us (we were also affected by the https://github.com/apache/maven-release/issues/1420 bug) so it feels like the release plugin isn’t really getting the care and attention it should, given how pivotal a plugin it is for Maven users (and presumably other Maven plugins?).
Note I have proposed a PR to revert this change - https://github.com/apache/maven-release/pull/1445
Note I’ve been around the ASF a fair while and I’m well aware that working on a tooling/library project like Maven is a fairly thankless task. Thank you for all the hard work you do to keep this awesome project going! I will be trying to keep an eye on this mailing list going forward and aim to test and vote on plugin releases where possible so that issues like this can hopefully be discussed before they make it out into the wild rather than after the fact in future.
Best Regards,
Rob Vesse
Re: CI Skip Breaking Change in maven-release-plugin 3.3.0
Posted by Konrad Windszus <kw...@apache.org>.
Hi Tamas,
> On 9. Dec 2025, at 12:10, Tamás Cservenák <ta...@cservenak.net> wrote:
>
> Hey Rob,
>
> sorry to hear this. My proposal would be to merge your rollback PR,
> and do a 3.3.1 release ASAP.
>
> Personally I am even surprised why the questioned PR got in in the
> first place, given it changes the _default value of an already
> configurable field_.
You were explicitly added as reviewer on https://github.com/apache/maven-release/pull/1423.
Also I consider the release notes quite verbose (https://github.com/apache/maven-release/releases/tag/maven-release-3.3.0)
> Breaking changes
• Include "[ci skip]" by default in scmReleaseCommitComment (#1423) @kwin
Still everyone voted +1 (https://lists.apache.org/thread/0wl7mkqgsbsxj9h85jxtgn0jg67bvsk1) including you.
Not sure why you are surprised by this….
> This is not how we usually handle such changes. Sorry for the inconvenience!
What other process do you propose except for maybe raising this in advance on the mailing list?
Thanks,
Konrad
>
> Thanks
> T
>
> On Tue, Dec 9, 2025 at 11:33 AM Rob @ DNR <rv...@dotnetrdf.org> wrote:
>>
>> Hey Maven folks,
>>
>> I wanted to raise concerns about a breaking change made in the 3.3.0 release of the maven-release-plugin
>>
>> This change was suggested in https://github.com/apache/maven-release/issues/1422 and provided in PR https://github.com/apache/maven-release/pull/1423
>>
>> Despite several well-founded objections to this idea being made in the original issue, and the PR, these were dismissed by the OP (a Maven PMC member AFAICT) and this change was merged with seemingly minimal discussion or consideration of its impacts on the many developers who depend on Maven for their builds.
>>
>> If you weren’t already aware this breaking change adds the "[ci skip]” text snippet into the commit message that the release plugin makes for the release commit, i.e. the SCM commit that has the non-SNAPSHOT version set and is tagged with the release tag. The effect of this text is that most common CI/CD systems ignore this commit and don’t build it.
>> However, many developers and organisations, myself included, have designed their CI/CD workflows around triggering release workflows from the release commit which this change breaks. Forcing us either to change our plugin configuration back to the old commit message default, or rollback to an earlier plugin version.
>>
>> Why was a breaking change of this magnitude not raised on the developers list and discussed more widely?
>>
>> This is in fact the second maven-release-plugin release in a row that has been unusable for us (we were also affected by the https://github.com/apache/maven-release/issues/1420 bug) so it feels like the release plugin isn’t really getting the care and attention it should, given how pivotal a plugin it is for Maven users (and presumably other Maven plugins?).
>>
>> Note I have proposed a PR to revert this change - https://github.com/apache/maven-release/pull/1445
>>
>> Note I’ve been around the ASF a fair while and I’m well aware that working on a tooling/library project like Maven is a fairly thankless task. Thank you for all the hard work you do to keep this awesome project going! I will be trying to keep an eye on this mailing list going forward and aim to test and vote on plugin releases where possible so that issues like this can hopefully be discussed before they make it out into the wild rather than after the fact in future.
>>
>> Best Regards,
>>
>> Rob Vesse
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> For additional commands, e-mail: dev-help@maven.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: CI Skip Breaking Change in maven-release-plugin 3.3.0
Posted by Tamás Cservenák <ta...@cservenak.net>.
Hey Rob,
sorry to hear this. My proposal would be to merge your rollback PR,
and do a 3.3.1 release ASAP.
Personally I am even surprised why the questioned PR got in in the
first place, given it changes the _default value of an already
configurable field_.
This is not how we usually handle such changes. Sorry for the inconvenience!
Thanks
T
On Tue, Dec 9, 2025 at 11:33 AM Rob @ DNR <rv...@dotnetrdf.org> wrote:
>
> Hey Maven folks,
>
> I wanted to raise concerns about a breaking change made in the 3.3.0 release of the maven-release-plugin
>
> This change was suggested in https://github.com/apache/maven-release/issues/1422 and provided in PR https://github.com/apache/maven-release/pull/1423
>
> Despite several well-founded objections to this idea being made in the original issue, and the PR, these were dismissed by the OP (a Maven PMC member AFAICT) and this change was merged with seemingly minimal discussion or consideration of its impacts on the many developers who depend on Maven for their builds.
>
> If you weren’t already aware this breaking change adds the "[ci skip]” text snippet into the commit message that the release plugin makes for the release commit, i.e. the SCM commit that has the non-SNAPSHOT version set and is tagged with the release tag. The effect of this text is that most common CI/CD systems ignore this commit and don’t build it.
> However, many developers and organisations, myself included, have designed their CI/CD workflows around triggering release workflows from the release commit which this change breaks. Forcing us either to change our plugin configuration back to the old commit message default, or rollback to an earlier plugin version.
>
> Why was a breaking change of this magnitude not raised on the developers list and discussed more widely?
>
> This is in fact the second maven-release-plugin release in a row that has been unusable for us (we were also affected by the https://github.com/apache/maven-release/issues/1420 bug) so it feels like the release plugin isn’t really getting the care and attention it should, given how pivotal a plugin it is for Maven users (and presumably other Maven plugins?).
>
> Note I have proposed a PR to revert this change - https://github.com/apache/maven-release/pull/1445
>
> Note I’ve been around the ASF a fair while and I’m well aware that working on a tooling/library project like Maven is a fairly thankless task. Thank you for all the hard work you do to keep this awesome project going! I will be trying to keep an eye on this mailing list going forward and aim to test and vote on plugin releases where possible so that issues like this can hopefully be discussed before they make it out into the wild rather than after the fact in future.
>
> Best Regards,
>
> Rob Vesse
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
For additional commands, e-mail: dev-help@maven.apache.org
Re: CI Skip Breaking Change in maven-release-plugin 3.3.0
Posted by Konrad Windszus <kw...@apache.org>.
Hi Rob,
Let me directly comment, as I was the one introducing the change.
> On 9. Dec 2025, at 11:32, Rob @ DNR <rv...@dotnetrdf.org> wrote:
>
> Hey Maven folks,
>
> I wanted to raise concerns about a breaking change made in the 3.3.0 release of the maven-release-plugin
>
> This change was suggested in https://github.com/apache/maven-release/issues/1422 and provided in PR https://github.com/apache/maven-release/pull/1423
>
> Despite several well-founded objections to this idea being made in the original issue, and the PR, these were dismissed by the OP (a Maven PMC member AFAICT) and this change was merged with seemingly minimal discussion or consideration of its impacts on the many developers who depend on Maven for their builds.
This is not correct. The PR was merged after approval before any complaints were raised (look at the history and the timestamps).
>
> If you weren’t already aware this breaking change adds the "[ci skip]” text snippet into the commit message that the release plugin makes for the release commit, i.e. the SCM commit that has the non-SNAPSHOT version set and is tagged with the release tag. The effect of this text is that most common CI/CD systems ignore this commit and don’t build it.
> However, many developers and organisations, myself included, have designed their CI/CD workflows around triggering release workflows from the release commit which this change breaks. Forcing us either to change our plugin configuration back to the old commit message default, or rollback to an earlier plugin version.
I was aware that this may be potentially breaking therefore it also had the breaking label (therefore is listed as breaking change in https://github.com/apache/maven-release/releases/tag/maven-release-3.3.0). Compare with the discussion at https://github.com/apache/maven-release/pull/1423#pullrequestreview-3482145032
> Why was a breaking change of this magnitude not raised on the developers list and discussed more widely?
It was discussed within the PR, however the impact of this change was probably wrongly perceived by myself. Seems lots of different ways are out there how you leverage m-release-p with CI.
Look for example at https://github.com/maveniverse/parent?tab=readme-ov-file#release-from-action how this can be done in CI.
>
> This is in fact the second maven-release-plugin release in a row that has been unusable for us (we were also affected by the https://github.com/apache/maven-release/issues/1420 bug) so it feels like the release plugin isn’t really getting the care and attention it should, given how pivotal a plugin it is for Maven users (and presumably other Maven plugins?).
I am sorry to hear that, however please understand this is a community project where everyone can only spend a limited time only. Feel free to involve yourself earlier and also test staging candidates during release.
>
> Note I have proposed a PR to revert this change - https://github.com/apache/maven-release/pull/1445
I will go ahead and merge if there are no further concerns raised by anyone else.
>
> Note I’ve been around the ASF a fair while and I’m well aware that working on a tooling/library project like Maven is a fairly thankless task. Thank you for all the hard work you do to keep this awesome project going! I will be trying to keep an eye on this mailing list going forward and aim to test and vote on plugin releases where possible so that issues like this can hopefully be discussed before they make it out into the wild rather than after the fact in future.
Greatly appreciated.
Regards,
Konrad
>
> Best Regards,
>
> Rob Vesse
>