Harness@5 attempting to fill some invisible questions

[kennsippell]

Describe the bug

To Reproduce
moh-kenya/config-echis-2.0@12981-noverify-when-chp
TZ=Africa/Nairobi npx mocha --reporter spec --timeout 120000 --slow 5000 ./test/tasks/postnatal_care_service.spec.js

Scenario leads to:

questions have new classes like contains-ref-target pre-init and contains-ref-target which form-filler thinks are visible

console errors

no context element found /postnatal_care_service/newborn_details/child/data/_screening/_not_feeding 0
ERROR JSHandle@object

I'm working on a minimal repro

Environment:

• Regression when upgrading from 3.x to 5.x

[jkuester]

Just a quick follow up here to note that I am digging into the error message:

no context element found /postnatal_care_service/newborn_details/child/data/_screening/_not_feeding 0
ERROR JSHandle@object

I have successfully reproduced this running live on a CHT instance, so this is not endemic to cht-form or the test-harness. While the error does not seem to prevent the form from being submitted with the necessary data, I think it is noticeable to the user. I am pretty sure that it is causing read-only text boxes to apprear on the summary page:

I am still running down the exact cause of the issue (will report back when I have more info), but it seems likely to me that it is an Enketo bug (enketo/enketo#99 resulted in a similar error in the logs).

[jkuester]

Okay, I have been able to track this no context element found error back down to a problem with the base Enekto logic and I have logged: enketo/enketo#1377.

See that issue for more details, but TLDR is that there seems to be issues handling relevant expressions in a repeat with no entries when adding additional entries to a separate repeat group....

[mrjones-plip]

Thanks for chasing down the issue @jkuester ! Is this an issue we can fix in CHT Core with a patch to our enketo then? Another way of asking this is: what workarounds, if any, are there? Is it just "avoid this combo in your form" kinda thing?

[jkuester]

So, on the enketo ticket I noted a workaround that seemed viable for the basic repro case (just making the repeat group non-relevant when it is empty), but I was never able to get the to work with the postnatal_care_service form (perhaps because the relevant expression for the field in the empty repeat actually depends on the field that determines the repeat_count...).

If it turns out to be a simple fix in Enketo, we can definitely patch it in cht-core while we wait on a new version of Enekto. However, I have not determined how this can/should be fixed in enketo-core at this point. More investigation is needed. @kennsippell (et al) I have logged medic/cht-core#9684 to track this on the CHT side. It would be useful to have folks weigh in out on that issue regarding the priority/impact of this issue. Thanks!

[mrjones-plip]

Thanks a bunch @jkuester !