As the title says, I'm trying to deploy some code and metadata from a sandbox to another. The deployment compiles successfully, and the tests pass too, but the deployment fails with the "Code coverage error" message.
The weird thing is that no coverage errors are shown. Usually it displays to the user which class is missing coverage, and how many lines are covered by the tests that were run.
What can be causing this?
Note: both environments are on API 44 (Winter '19).
This is a single-language org, so the interface is in portuguese. But as you can see, no classes are listed.
Best Answer
I encountered the same issue as recently. We have an appexchange app that handles deployments to different orgs using the Metadata api. Since we upgraded we got support cases mentioning that the deployment was running all good, no test failures reported but there was a code coverage error without specifying how much percent is missing (what is the normal behavior). And even if the org had more than 75% apex coverage.
After digging deeper in the issue we found the following cause:
Since Winter'19, API version 44.0, Salesforce is also checking code coverage on flow as soon as they are deployed as active. You can find the information in these release notes:
https://releasenotes.docs.salesforce.com/en-us/winter19/release-notes/rn_forcecom_flow_test_coverage.htm
https://releasenotes.docs.salesforce.com/en-us/winter19/release-notes/rn_forcecom_flow_deploy_as_active.htm
However the release notes state that this should be activated on the org and if not active the coverage is not applied. Again some deeper investigation from our side and we found that the setting is only available on production orgs (default false) but not on sandboxes. So deployments to sandbox always seem to calculate code coverage on flow or process builders. This was causing the deployments to fail unexpectedly, we found the actual error message in the
FlowCoverageWarningsobject received fromcheckDeploymentStatusmethod in the metadata api.I'm not a 100% sure that what you encountered has the same cause, but I'm inclined to say it is.
I see only two options to avoid this in the future, either don't run tests deploying to sandbox (not really advisable) or provide test classes that cover process builder / flow logic so that it provides >75% on them.
Hope this helps you further!