Pipeline migration produces invalid container image reference in pipeline file

Describe the issue

Currently we have two hooks touching the same value (runtime image) which are expected to occur sequentially, but are actually triggered separately by changes to the jupyterlab widget content and the canvas content respectively. In theory this should be the same, but is causing this race condition when only the widget detects changes but the canvas does not.

To Reproduce Steps to reproduce the behavior:

1. Clone https://github.com/elyra-ai/examples

2. Open examples/pipelines/dax_noaa_weather_data/analyze_NOAA_weather_data.pipeline in a text editor and review all runtime_image property values:

   "runtime_image": "docker.io/amancevice/pandas:1.1.1"
   

3. Open examples/pipelines/dax_noaa_weather_data/analyze_NOAA_weather_data.pipeline in the VPE

4. The migration wizard opens. Migrate the pipeline.

5. Save the pipeline and close the pipeline editor.

6. Open the pipeline file again in a text editor and review the runtime_image property values:

   "runtime_image": "Pandas 1.1.1"
   

   Note that the updated values are not valid container image names.

7. Try to run the pipeline. It fails because there is no image with such a name.

Expected behavior

The two hooks content should either not be dependent on each other or we should make sure one hook triggers the next if that behavior is expected.

Deployment information Describe what you’ve deployed and how:

• Elyra version: 3.2/rc0

[ptitzler]

IIUC this will still work if using only the VPE since it converts it to the correct format at submission.

It doesn’t seem to run from the VPE either when you run the pipeline after the migration without doing anything else.

[ajbozarth]

I tracked down the issue and opened #2231 with a fix. The actual bug is described in detail there