This has happened quite a few times and seems to be when the diagram is updated from the server. I am not sure if the execution bar remains in place or the lifeline remains in place, but it always seemed that the execution bars were to the right of the lifeline.
If the lifeline is moved even slightly then the execution bars jump back into proper alignment.
I have the Linux version which I have run on both FreeBSD 12.2 and Ubuntu 20.04.1 LTS (focal) and seen the effect on both systems. I understand from aco-worker that this can happen on the Windows version.
I am the only one who has been modifying the project, but my colleague has accessed it several times. That particular diagram I think was created after our last meeting, so was probably only accessed by me. I think I only used the VM to both create it and read it back. However I did open, but not edit every other Sequence Diagram in the project, but not any of the Block Diagrams using the diagram switcher. I think I had closed the program, but the VM was left running while was doing other tasks.
I just noticed that the exported PNG had the execution bar off the lifeline, even though I had closed the program and re-opened it to change a different diagram. I had to switch to that diagram and move the lifeline to beable to export a correct image.
Thank you for the files and steps. Unfortunately, we are unable to repeat the problem. I have captured a video to show you our steps. Could you take a look? You can watch here:
By the end of the video, I opened the first revision and performed a commit. That action was to simulate the update of project change made in another machine (i.e. your step 6)
I was able to get replication on a single machine, but it was the “other” diagram that was affected, not the one was displayed.
My method was:
open VP
move a lifeline on whatever diagramwas currentlyselected
add a message involving that lifeline
Ctrl-Tab toggle to second diagram
move a lifeline
add a message involving that lifeline
close, save and commit
It took 6 repetitions to see the effect
(actually I was trying to repeat the lost description)
Currently I think multiple diagrams have to be changed before the effect shows
I have a screen capture of the effect - it took me over 30 min to see the effect happen. However I could upload just the last5 min to show a couple of the cycles if that might help
I uploaded the clip which is just the last couple of minutes showing the bug appear. The full file is just over 30 min and is just me adding and removind arrows and lifelines and closing/opening the VP program watching for the bug to happen.
A rather boring video, Ithink
This is recorded with the program running in an Ubuntu 20.04.1 in a VM on a FreeBSD host.
With your help, we can repeat the problem now. Thank you very, very much! The problem seems to be related to two moves: (1) move a sequence message vertically in a way that causes the target activation to be detached from the original bar of activation and (2) Move the lifeline
I have already asked our engineers to check and fix. Once again, thank you so much for your help.
I am glad to inform you that the bug has been fixed. You can now update to the latest patch build (16_2_20210210ac or higher) to obtain the fix. For instructions, please read: