Hi all,
I have a very strange issue that has been causing a lot of issues in production. We automate a desktop app which purchases batches (essentially invoices). If any error occurs during the processing of the batch, the bot will close the batch. Previously, this would cause the batch to be saved and the bot would flag the batch for human review through another system. We recently updated the application which handles batches and now there is a save dialog that appears when closing the batch.
I have created a custom sequence to handle this prompt, both because this can appear in multiple places and because this creates an easier time testing. This sequence has no input parameters.
When running this sequence by itself, regardless of debug (with breakpoints) or production builds, environment, or server the code runs on, it will consistently identify and click the Yes button on the prompt as intended. However, when this sequence is plugged in to the rest of the bot and executed, the bot will consistently fail to identify the prompt exists, even though it is definitely there.
I have tried:
- Legacy
- Element Exists
- Get Text targeting the body
- Get Text of the OK button
- Modern
- Check App State
- Strict Selector
- Fuzzy Selector
- Image Selector
- Despite being able to select the native text option in the selector, this does not appear to save.
- I am not licensed for CV activities, so these have not been attempted.
- Check App State
Across hundreds of batches, I have had a 0% success rate in the bot identifying the prompt when running the close save dialog sequence with the rest of the bot. Logging indicates that it successfully enters the sequence, but consistently takes the “prompt is not present” code branch.
What is different about these two scenarios? Why would I have a 100% success rate when running this sequence in isolation but a 0% success rate when running through an Invoke Workflow activity?