SMS_R_Ld5.ne4_ne4.FSCM5A97 test fails with PIO2 #2963
Comments
|
This test fails with the default PIO1 on anvil as well: This test has either never passed on anvil or hasn't passed in a long time. |
|
I found the issue. ROOTPE needs to be explicitly set to 0 for all components in cime/scripts/lib/CIME/test_scheduler.py for PTS_MODE. The test worked on anvil once I did this. I will issue a cime PR to address this. |
|
will do |
|
@jayeshkrishna are you talking about including the mod to file components/cam/src/control/ncdio_atm.F90 ? If so, I can't include that in my cime PR since it's not in the cime/ directory |
|
Ok, in that case please don't close this issue (Refer to the issue "See Issue #...") when you merge your PR. I will create a separate PR with the PIO fix and merge it (closing this issue) |
|
@bogensch if you're making the PR here, you can mix cam and cime file changes. Just don't put them in the same commit. |
The SMS_R_Ld5.ne4_ne4.FSCM5A97 test was failing on Anvil. The reason is that ROOTPE needs to be explicitly set to zero for all component models for PTS_MODE (single column model mode). All developer tests pass with this change. Addresses #2963 [BFB]
The SMS_R_Ld5.ne4_ne4.FSCM5A97 test was failing on Anvil. The reason is that ROOTPE needs to be explicitly set to zero for all component models for PTS_MODE (single column model mode). All developer tests pass with this change. Addresses #2963 [BFB]
Remove support for CLM4.0 CLM4.0 is being dropped in CESM, and has already been dropped in E3SM. This commit removes some of the complexity related to supporting that old version of CLM. This commit completely removes support for CLM4.0 for CESM. For E3SM, support is removed in files that are shared between CESM and E3SM, but I have not made any changes to E3SM-specific files. A key change was in the build: we no longer need to distinguish between CLM4.0 vs. later versions of CLM in setting some build options (but I think that we do still need to determine whether we're running with CLM or some other land model, including dlnd/slnd/xlnd, so some CLM-specific logic still remains in the Makefile). This change is backwards-incompatible with respect to CLM4.0 support, but should be backwards compatible for cases using later versions of CLM, as well as cases using other land components. Test suite: scripts_regression_tests on cheyenne Also ran most of the aux_clm test suite in the context of ESCOMP/CTSM#609 (note that that used a version of this cime branch rebased onto cime5.7.5) Test baseline: For aux_clm test suite, used previous ctsm tag Test namelist changes: none Test status: bit for bit Fixes #2963 User interface changes?: Drops support for CLM4.0 Update gh-pages html (Y/N)?: N Code review:jedwards
CAM uses an invalid PIO file id while inquiring info about a NetCDF file.
The error message is:
[0] ERROR: (shr_scam_getCloseLatLonNC) ERROR from nf90_inquireThe bug is at components/cam/src/control/ncdio_atm.F90:205
call shr_scam_getCloseLatLon(ncid%fh,scmlat,scmlon,closelat,closelon,latidx,lonidx)For PIO2, the ncid returned to the user is not the actual ncid returned by the underlying netcdf/pnetcdf library. This opaque pio_ncid is unusable from outside PIO2, such as nf90_inquire calls (direct calls to the NetCDF library, bypassing PIO) in shr_scam_getCloseLatLonNC.
PR #2771 has fixed similar issues for other shr_scam_getCloseLatLon calls. However, the fixes merged in the PR were created before line 205 (with the bug) was added to ncdio_atm.F90, so the PR did not end up fixing this bug.
The text was updated successfully, but these errors were encountered: