Merged in fixup/merge_803_into_next (pull request #252)
d8aabb7·Author: David Dickinson·Closed by: Joseph Parker·2020-02-10
Description
Merge the 8.0.3 release into next, fixing conflicts. Replaces discussion PR #251
Make sure we respect read-only petsc vectors
Update doi in README
Note: this is now the "all versions" doi, rather than the "latest version". The all versions link resolves to the latest version on Zenodo, but the link won't need updating when new versions are released.
Add explicit targets to the generated source files
This helps avoid issues when the dependency file exists but the $(GENERATED_SRC_DIR) doesn't. For example if a user has done make clean cleandirs, prior to this patch a subsequent make gs2 would fail (if not using ONE_STEP_PP).
Avoid requiring SLEPC_ARCH to be set for eigensolver builds.
This is a non-standard flag that is used where PETSC_ARCH should really be used. In the future we can remove references to SLEPC_ARCH, but we keep for now to maintain compatibility.
Bug fix for fortdep -- allow it to recognise abstract interface
gs2_save.fpp edited online with Bitbucket to fix bug in gs2_save -- restart files incorrectly tried to save and restore the last but two timestep size, resulting in potentially undefined coefficients for restarted runs.
Allow OpenMPI to oversubscribe on GitLab CI
Make smart default for write_fluxes
Ensure we increment nout before leaving the diagnostics routine
This fixes a bug with runs that disable the majority of the diagnostic output. The time dimension was remaining length 1, meaning the always present time dependent data only ever contained the last time point.
Pull out formatting of git log into separate function
Generalise the changelog generator to work between any two commits
Fix Jenkins checkout due to update of plugin
Minor bug fix, ensures that building auxiliary programs (such as dump_response etc.) will create bin (or more strictly $(BINDIR) ) if it doesn’t exist.
Approved-by: Peter Hill peter.hill@york.ac.uk Approved-by: David Dickinson d.dickinson@york.ac.uk (cherry picked from commit 795a1610f7d4546e6cd5d562a265b73df526a3dd)
Ensure exit_reason file is removed before test
Manipulate exit_reason file on proc0 only
Allow checkstop to manipulate state%exit_code
Provide exit code for exiting normally on completing nsteps
Fix style
Make coment FORD style
Append, not overwirte, the exit_reason file
Change EXIT_NO_REASON to EXIT_NOT_REQUESTED
Merged in bugfix/fix_jenkins_netcdf_dependency_changing (pull request #238)
Try switching netcdf source to github from unidata ftp
Try switching netcdf source to github from unidata ftp
Fix directory names
Remove erroneous v
Fix other builds
Try updating to the latest netcdf versions
Revert "Try updating to the latest netcdf versions"
Merge the 8.0.3 release into next, fixing conflicts. Replaces discussion PR #251
Make sure we respect read-only petsc vectors
Update doi in README
Note: this is now the "all versions" doi, rather than the "latest version". The all versions link resolves to the latest version on Zenodo, but the link won't need updating when new versions are released.
Add explicit targets to the generated source files
This helps avoid issues when the dependency file exists but the
$(GENERATED_SRC_DIR)
doesn't. For example if a user has donemake clean cleandirs
, prior to this patch a subsequentmake gs2
would fail (if not usingONE_STEP_PP
).Avoid requiring SLEPC_ARCH to be set for eigensolver builds.
This is a non-standard flag that is used where PETSC_ARCH should really be used. In the future we can remove references to SLEPC_ARCH, but we keep for now to maintain compatibility.
Bug fix for fortdep -- allow it to recognise abstract interface
gs2_save.fpp edited online with Bitbucket to fix bug in gs2_save -- restart files incorrectly tried to save and restore the last but two timestep size, resulting in potentially undefined coefficients for restarted runs.
Allow OpenMPI to oversubscribe on GitLab CI
Make smart default for write_fluxes
Ensure we increment nout before leaving the diagnostics routine
This fixes a bug with runs that disable the majority of the diagnostic output. The time dimension was remaining length 1, meaning the always present time dependent data only ever contained the last time point.
Pull out formatting of git log into separate function
Generalise the changelog generator to work between any two commits
Fix Jenkins checkout due to update of plugin
Minor bug fix, ensures that building auxiliary programs (such as dump_response etc.) will create bin (or more strictly
$(BINDIR)
) if it doesn’t exist.Bug fix for dump_response program
Ensure dump_response has the fft dependency
Use default exit reason
Merged in bugfix/stop-file-test-exit-code (pull request #204)
Use exit code in job manage test
Approved-by: Peter Hill peter.hill@york.ac.uk Approved-by: David Dickinson d.dickinson@york.ac.uk (cherry picked from commit 795a1610f7d4546e6cd5d562a265b73df526a3dd)
Ensure exit_reason file is removed before test
Manipulate exit_reason file on proc0 only
Allow checkstop to manipulate state%exit_code
Provide exit code for exiting normally on completing nsteps
Fix style
Make coment FORD style
Append, not overwirte, the exit_reason file
Change EXIT_NO_REASON to EXIT_NOT_REQUESTED
Merged in bugfix/fix_jenkins_netcdf_dependency_changing (pull request #238)
Try switching netcdf source to github from unidata ftp
Try switching netcdf source to github from unidata ftp
Fix directory names
Remove erroneous v
Fix other builds
Try updating to the latest netcdf versions
Revert "Try updating to the latest netcdf versions"
This reverts commit fd3f8de2261e1528d8a46475ae4ccdbd067af74e.
Requires changes to Jenkins makefile etc. which we can reserve for a more significant release.
Approved-by: Joseph Parker joseph.parker@stfc.ac.uk Approved-by: David Dickinson d.dickinson@york.ac.uk
Add smart default for write_nl_flux in old diagnostics
Don't try to specify the exit_reason syntax
Update the utils version
Move test to after simulation finished
Replace
system
withexecute_command_line
in testsRename theta_grid:eps to theta_grid:eps_trapped to distinguish this from the related theta_grid_params:eps used by the s-alpha model
Update netcdf output with eps_trapped
Add long_name attribute for eps in netcdf output
Update new diagnostics for change of name from eps to eps_trapped
Fix typo in attribute
skip ci
Correct typo "CPP Plasma" to "CCP Plasma"
Reorder statements to ensure we set variables based on the input values rather than the default values
Both
fft_use_wisdom
andfft_measure_plan
are set in theinit_gs2_layouts
callBackport from PR #164
Update for changes in utils
Update 8.0.3-RC latest versions of externals
This commit updates utils to include:
bugfix/single-precision-timing (pull request #30)
feature/write-exit-reason (pull request #32)
bugfix/guard_against_empty_send_in_redistribute (pull request #31)
and Makefiles to include:
bugfix/marconi-hdf5 (pull request #30)
bugfix/archer-build_gs2 (pull request #31)
fedora_makefile_parallelnetcdf_fix (pull request #29)
feature/store-verbosity-backwards-compatible (pull request #36)
feature/update-exit-codes (pull request #33)
Update CITATION.cff and README for 8.0.3
Update to release versions of utils and Makefiles
Fix job manage test for changes in exit_code
Update utils again to include the updated changelog
Update CHANGELOG.md
Update changelog
Put 8.0.3 in RELEASE file and make sure Makefile parses this as a string
Release date: 10/02/20
Â