fix(api): restore empty error blocks on cancelled runs #15215
Conversation
There was a problem hiding this comment.
Whoof, nice find, sorry about that.
The tests show this failing some E-stop-related test that I added in #14929. I'm trying to remind myself of its implications.
There's certainly a way to solve both problems at once, but I'm a little reluctant to do more PE state work this late in v7.3. What do you think?
|
Wow subtle indeed, this fixed the issue. |
The implication would be that, conversely, an estop error would not get reported. 4c30bf5 i'm pretty sure restores the exact thing your PR was meant to remove but such is life on a timeline, and I think it will work. |
There was a problem hiding this comment.
OK, sounds like a plan, presuming it passes the on-robot tests. (Thank you @vegano1 for testing!)
|
The change is working as expected, see attached images.
|
Thank you! |
I think this would be a subtle side effect of a previous pr trying to improve PE end behavior.
The reason this fix is required is that when a client cancels a protocol by stopping the engine, the
StopActionsets the run result (good) and doesn't set a run error (good, correct, there wasn't one, this is a cancel). In 7.2, with the behavior this PR now restores, aFinishActionthat might contain an error wouldn't set that error in the run if it already had a result, whether or not it had an error. In 7.3, it would set the error if the engine had no error, which it wouldn't, because there is no error when you stop.Testing
Closes RQA-2737