One failure mode of the e2e-tests which I see quite frequently is that the text entered to confirm the local password is much longer than it's supposed to be:
This happened especially often in this run, where it caused five of the six failures, but I also see it in other test runs.
Looking at the logs, I see two issues:
- It seems like the "Return" key which should be pressed here is lost somehow.
- The test continues even though this Match Text call should fail, but it doesn't.
Regarding 1., I've seen the Return key being ignored in other cases, but it's strange that it seems to happen much more often in this specific case. A solution could be to check that the "Confirm password:" text is not there anymore, and if it is, press Return again. But if we have to do this kind of check in all cases where we use the Return key, or keyboard input at all, it's a lot of overhead.
Regarding 2., I added commit to #1412 to log the image where it found the text, and the text it read with confidence and similarity. That should help debug false positives like that.
One failure mode of the e2e-tests which I see quite frequently is that the text entered to confirm the local password is much longer than it's supposed to be:
This happened especially often in this run, where it caused five of the six failures, but I also see it in other test runs.
Looking at the logs, I see two issues:
Regarding 1., I've seen the Return key being ignored in other cases, but it's strange that it seems to happen much more often in this specific case. A solution could be to check that the "Confirm password:" text is not there anymore, and if it is, press Return again. But if we have to do this kind of check in all cases where we use the Return key, or keyboard input at all, it's a lot of overhead.
Regarding 2., I added commit to #1412 to log the image where it found the text, and the text it read with confidence and similarity. That should help debug false positives like that.