You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[JENKINS-24589] Build History - name and links overlap if build names are long #23426
I've been seeing this as well. I've tried setting the names to something as short (and still descriptive) as possible, but after upgrading to 590 I'm still seeing the overlap. I use single jobs to build from multiple branches of the same product with a branch monicker in the buildname (via build name setter plugin) as a way to differentiate them for my users.
I've been seeing this as well. I've tried setting the names to something as short (and still descriptive) as possible, but after upgrading to 590 I'm still seeing the overlap. I use single jobs to build from multiple branches of the same product with a branch monicker in the buildname (via build name setter plugin) as a way to differentiate them for my users.
Yes I get this behaviour with post build annotation we put into description. It looks really bad and unprofessional.
Comments I frequently get now from Senior management, are "Why don't you get a professionally created tool"?
With this on the UI, it looks like a typical "hacky Open Source" project.
I don't know why anyone bothered to update the UI.
Yes I get this behaviour with post build annotation we put into description. It looks really bad and unprofessional.
Comments I frequently get now from Senior management, are "Why don't you get a professionally created tool"?
With this on the UI, it looks like a typical "hacky Open Source" project.
I don't know why anyone bothered to update the UI.
IMO this should be removed from .3. It is not suitable for LTS, as it appears to break HTML descriptions (JENKINS-26406) and has a performance regression when revealing a few hundred builds by clicking 'more' (JENKINS-26445). It's also no longer possible to click the ball to go to the build output page.
A post on the dev list seems to indicate that .3 introduced an issue that may be JENKINS-25971, whose fix (it's actually fixed) caused a few more regressions (JENKINS-26312JENKINS-26298JENKINS-26306 I think) that were finally fixed in 1.596 or so.
IMO this should be removed from .3. It is not suitable for LTS, as it appears to break HTML descriptions (JENKINS-26406) and has a performance regression when revealing a few hundred builds by clicking 'more' (JENKINS-26445). It's also no longer possible to click the ball to go to the build output page.
A post on the dev list seems to indicate that .3 introduced an issue that may be JENKINS-25971, whose fix (it's actually fixed) caused a few more regressions (JENKINS-26312JENKINS-26298JENKINS-26306 I think) that were finally fixed in 1.596 or so.
I would check the conclusions Daniel has come to carefully, I think I have both the issues he indicates.
However, they only seemed to be present for me when upgrading to 1.586.
This change may not have introduced these issues.
The improvement with non-overlapping, I would not like to see regressed.
I would check the conclusions Daniel has come to carefully, I think I have both the issues he indicates.
However, they only seemed to be present for me when upgrading to 1.586.
This change may not have introduced these issues.
The improvement with non-overlapping, I would not like to see regressed.
The recent discussion is about backporting this change to the LTS release 1.580.3 only (which has been rejected due to the regressions this causes). It has nothing to do with the weekly releases which will continue to include this new build history.
I admit my comment wasn't too clear when only viewing comments to this issue, as there has been no discussion about LTS inclusion, only changes to the issue tags indicated that.
The recent discussion is about backporting this change to the LTS release 1.580.3 only (which has been rejected due to the regressions this causes). It has nothing to do with the weekly releases which will continue to include this new build history.
I admit my comment wasn't too clear when only viewing comments to this issue, as there has been no discussion about LTS inclusion, only changes to the issue tags indicated that.
Activity
jenkins-infra-bot commentedon Sep 4, 2014
Raw content of original comment:
Duplicates
JENKINS-24589.jenkins-infra-bot commentedon Sep 4, 2014
Raw content of original comment:
Correction,
JENKINS-23829was about different behavior and is actually the cause of this.jenkins-infra-bot commentedon Sep 4, 2014
Raw content of original comment:
Yes we see this as well now.. granted the UI is much better.. but hard to get to the results now for some jobs
jenkins-infra-bot commentedon Sep 4, 2014
Raw content of original comment:
Screenshots, if sufficiently different from the reported ones, may be useful to determine how to best improve the situation.
jenkins-infra-bot commentedon Sep 5, 2014
Raw content of original comment:
There was an intention to raise a special "collector-ticket" for this. Linking to Jenkins-24353.
jenkins-infra-bot commentedon Sep 11, 2014
Raw content of original comment:
In 579 it can look like this attachement:
jenkins-infra-bot commentedon Nov 21, 2014
Raw content of original comment:
I've been seeing this as well. I've tried setting the names to something as short (and still descriptive) as possible, but after upgrading to 590 I'm still seeing the overlap. I use single jobs to build from multiple branches of the same product with a branch monicker in the buildname (via build name setter plugin) as a way to differentiate them for my users.
jenkins-infra-bot commentedon Nov 21, 2014
martouf:
Raw content of original comment:
Yes I get this behaviour with post build annotation we put into description. It looks really bad and unprofessional.
Comments I frequently get now from Senior management, are "Why don't you get a professionally created tool"?
With this on the UI, it looks like a typical "hacky Open Source" project.
I don't know why anyone bothered to update the UI.
jenkins-infra-bot commentedon Nov 21, 2014
Raw content of original comment:
Some related discussion and suggestions at #1466
Please consider commenting how/why/when a proposed solution would (not) work well. (Please limit your comments to constructive criticism.)
I'll try to ask tfennelly in IRC about a short-term solution suitable for 1.580.x.
jenkins-infra-bot commentedon Nov 25, 2014
Raw content of original comment:
Please see #1470. Try it if you can and provide feedback.
jenkins-infra-bot commentedon Nov 26, 2014
Raw content of original comment:
I implemented the war file from pull 1470 and it did clean it up a bit. The text no longer overlaps and is easier to read. Thank you!
jenkins-infra-bot commentedon Nov 26, 2014
Raw content of original comment:
@Michael thanks for trying PR #1470. Sounds as though it did not clean it up quite to your satisfaction though, right?
jenkins-infra-bot commentedon Nov 30, 2014
Raw content of original comment:
Fixed in cb9036c
jenkins-infra-bot commentedon Jan 20, 2015
Raw content of original comment:
IMO this should be removed from .3. It is not suitable for LTS, as it appears to break HTML descriptions (
JENKINS-26406) and has a performance regression when revealing a few hundred builds by clicking 'more' (JENKINS-26445). It's also no longer possible to click the ball to go to the build output page.A post on the dev list seems to indicate that .3 introduced an issue that may be
JENKINS-25971, whose fix (it's actually fixed) caused a few more regressions (JENKINS-26312JENKINS-26298JENKINS-26306I think) that were finally fixed in 1.596 or so.jenkins-infra-bot commentedon Jan 21, 2015
Raw content of original comment:
@Daniel sorry, what exactly do you mean when you say ".3" ?
I'd prefer to try fix these issues (when I get a chance) Vs going back to the way it was before, with labels etc overlapping.
jenkins-infra-bot commentedon Jan 21, 2015
Raw content of original comment:
Ah ... you mean 1.580.3 I think.
jenkins-infra-bot commentedon Jan 21, 2015
martouf:
Raw content of original comment:
I would check the conclusions Daniel has come to carefully, I think I have both the issues he indicates.
However, they only seemed to be present for me when upgrading to 1.586.
This change may not have introduced these issues.
The improvement with non-overlapping, I would not like to see regressed.
jenkins-infra-bot commentedon Jan 21, 2015
Raw content of original comment:
The recent discussion is about backporting this change to the LTS release 1.580.3 only (which has been rejected due to the regressions this causes). It has nothing to do with the weekly releases which will continue to include this new build history.
I admit my comment wasn't too clear when only viewing comments to this issue, as there has been no discussion about LTS inclusion, only changes to the issue tags indicated that.
jenkins-infra-bot commentedon Nov 24, 2025
[Original
is_duplicated_byfrom Jira: JENKINS-24687]jenkins-infra-bot commentedon Nov 24, 2025
[Original
is_duplicated_byfrom Jira: JENKINS-24619]