The diffs described in #3825 got me worried that there may have been other diffs introduced in the NaN-fixing work—tags ctsm5.4.022 (#3787) and ctsm5.4.023 (#3809)—that we didn't notice because the tests had been failing with the ifx compiler. To check, each of those tags should be tested using the old intel compiler (or gnu), expecting no diffs relative to ctsm5.4.021, for each of the following suites:
aux_clmfatesctsm_scicrop_calendars
This will first require generating ctsm5.4.021 baselines for the tests that aren't in aux_clm, I think. Using --extra-create-test-args " --skip-tests-with-existing-baselines" will make it so each test only runs once.
Marking as high priority and "next" so we can make sure this is planned soon, since this could reveal answer changes that need to be reverted.
The diffs described in #3825 got me worried that there may have been other diffs introduced in the NaN-fixing work—tags ctsm5.4.022 (#3787) and ctsm5.4.023 (#3809)—that we didn't notice because the tests had been failing with the ifx compiler. To check, each of those tags should be tested using the old intel compiler (or
gnu), expecting no diffs relative to ctsm5.4.021, for each of the following suites:aux_clmfatesctsm_scicrop_calendarsThis will first require generating ctsm5.4.021 baselines for the tests that aren't in
aux_clm, I think. Using--extra-create-test-args " --skip-tests-with-existing-baselines"will make it so each test only runs once.Marking as high priority and "next" so we can make sure this is planned soon, since this could reveal answer changes that need to be reverted.