-
|
Hi, today i went into some strange trouble using the native Jenkins support for Windows. After some updates of our infrastructure the previously build and uploaded artifacts are no more in sync. If I try to build the same state of the recipes again on Jenkins the Query fails for all packages. I am using Bob 0.25.0 but the same happens with an older version of bob (like 0.23.1). I also used the
The left side is the one of my local machine and the right side of the Jenkins. All other values are the same. Does this attribute have an effect on the Build-Id? Does anyone have any ideas or can anyone give me any hints on how I can further isolate the problem? Pythonversion: 3.11.7 |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments
-
|
This sounds like something has changed that influences the hash sum of the sources. But this is just a wild first guess. I think the most promising approach would be to compare the audit trails of identical packages from Jenkins and your local builds. Preferably, this is done with some package that has as less dependencies as possible but that still shows the problem. The difference in your One Windows thing that comes to my mind is the support of symlinks. On Windows, Bob makes a fundamental distinction in binary artifacts whether the host supports symlinks or not. So if your Jenkins server does not has symlink support enabled, but your local machine does, then they won't share artifacts. See |
Beta Was this translation helpful? Give feedback.
-
|
I have checked it briefly and indeed, a “wrong” version of Bob was deployed on the Jenkins. Furthermore, the support for symlinks was overwritten for some reason and is therefore no longer set. I will check the issue tomorrow, but I think that is the root cause of the problem. Thank you very much for the quick help and the good advice. |
Beta Was this translation helpful? Give feedback.
-
|
The root-cause was the lack of authorization in Jenkins for setting symlinks. |
Beta Was this translation helpful? Give feedback.
This sounds like something has changed that influences the hash sum of the sources. But this is just a wild first guess. I think the most promising approach would be to compare the audit trails of identical packages from Jenkins and your local builds. Preferably, this is done with some package that has as less dependencies as possible but that still shows the problem.
The difference in your
bob showcommand looks like 8502cf9 but that is not on any release yet. So the right side of your diff would be current Bob master branch. Otherwise I don't see how that could happen. In any case, this attribute has no influence on the Build-Id. So it cannot explain the behaviour.One Windows thing tha…