Testing: Disable 2018 answer flags #1485
Conversation
The TC tests are still using 2018 answer reproducibility flags, which fail when aggressive initialization of allocatable arrays is enabled. In at least one case, the `mask_z` variable of the ALE sponge is incompletely initialized when `DEFAULT_2018_ANSWERS` is set. When unset, the array is fully initialized. This patch disables the 2018 answer flags and restores reproducibility on such platforms.
marshallward
force-pushed
the
testing_no_2018
branch
from
8b16a2c to
f347b10
Compare
Codecov Report
@@ Coverage Diff @@
## dev/gfdl #1485 +/- ##
============================================
- Coverage 29.13% 29.08% -0.05%
============================================
Files 235 235
Lines 71097 71097
============================================
- Hits 20715 20681 -34
- Misses 50382 50416 +34
Continue to review full report at Codecov.
|
There was a problem hiding this comment.
These changes shift the TC testing to emphasize the newer code that we want people to be using, although it does not preclude the use of the older algorithms in existing experiments, which will be retained as there is any demand to use them. I agree with these changes.
The TC tests are still using 2018 answer reproducibility flags, which
fail when aggressive initialization of allocatable arrays is enabled.
In at least one case, the
mask_zvariable of the ALE sponge isincompletely initialized when
DEFAULT_2018_ANSWERSis set. Whenunset, the array is fully initialized.
This patch disables the 2018 answer flags and restores reproducibility
on such platforms.