Modeler creating invalid BPMN20...

[wolpert]

I've got an example of where modeler is creating a BPMN20 xml file that should be valid by design, but fails. The rough steps are build something good, make it bad, try to fix it. You can't fix it it seems.  Here they are specifically…

1) Create a (valid) model with nested xor gateways in modeler. (Importing model via Probe works)
2) Modify model so that the xor gateways are logically invalid. (Importing model via Probe fails)
3) Fix model by removing changes, so it looks good again, not by hitting 'undo'. (Importing model still fails, when the gui 'looks' like the original.

I've done this example, and have an included tar file with resulting xml files. Step one files are in the 'Working' directory. Step 3 files are in the 'NotWorking' directory.  The files from step '2' are in the directory named Intermediary. If modeler is open during the whole process… you can use 'undo' to go back in time to step 1 when it works. This obviously doesn't help if a diagram took multiple days, where the app may have been restarted.

[wolpert]

I have a simpler example of the above. I moved one task visually over to the left by a small amount. Resulting in a change to the xml file that will not be loaded via Probe into activiti. So its one step different. I have no idea if I'm doing something wrong…  When it fails, I get the following error (in this case)

Invalid destination 'sid-4C583F4A-15B1-4928-91AF-B2B1265620B4' of sequence flow 'sid-934201BF-BA5C-4299-8B7C-5299B4735198' | TestDeploy.bpmn20.xml | line 28 | column 182
   Invalid source 'sid-BB4DE8CB-D0D6-49E2-BA4B-3DD4CB870A72' of sequence flow 'sid-E8487D13-D77C-4ACD-A78B-59D02EEBED4B' | TestDeploy.bpmn20.xml | line 29 | column 182

Keep in mind I'm deploying the file created by the modeler… not exporting it via cycle. Is that a valid thing to do?

[wolpert]

Just wanted to see if anyone else can confirm this problem by using the included files in modeler's file directory and importing them via probe. I'd like to know if some issues with modeler like this have been seen before.

Also, the two xml files are so different considering the only thing I did in modeler was move one task < 1cm to the right… it decided to change SIDs in the process. Is that normal and expected?

[jbarrez]

This is definitely not to be expected.

I tried to mimic what you've described: I created a simple process (start -> task -> end). Shut down tomcat with Modeler. Rebooted and moved task around and tried to reupload. That seems to work every time.

Can you explain with a very simple example how exactly you get the error? Which version are you using?

[wolpert]

You should be able to view the second tar file… copy the contents to your workspace and open via modeler. Specifically, its the case of nested conditional that I think exposes the bug, but that may be just how I achieve it.

In the second tar file, there are two variants. The first is the state I ended up with that worked, the second was moving one task over to the right, visually.

I got the original state by creating a process with one conditional, and along one resulting path, a second conditional. I then 'crossed' streams of that second condition to the first. Then back to normal. that was able to be imported. Then moving tasks around caused failures after that point.

Let me know if the tar file is useful, or if not, I can try to make other examples. I'm able to make this fail pretty regularly; such that I have to backup and use the eclipse designer instead.