Releases: Goddard-Fortran-Ecosystem/pFUnit
Releases · Goddard-Fortran-Ecosystem/pFUnit
Update minimum required CMake version
GFortran workaround
Fixed
- Workaround for gfortran 13/14 on Ubuntu. Failure does not show on other fla
vors of Linux no macos.
Misc updates (support for Flang)
Changed
- Updated submodule for fArgParse
- Update CI to have
gfortran-10
andgfortran-11
only onubuntu-22.04
- Update CI NVIDIA to NVHPC 24.7
- Add Flang to CI
Migrated to use v2 interfaces from gFTL
Changed
- Updated pFUnit to use v2 template interfaces
- Added
-quiet
flag for NAG Fortran - Remove
macos-11
from GitHub Actions, addmacos-12
andgfortran-14
to Ubuntu 24.04 - Updated fArgParse to v1.8.0
Added
- LLVMFlang compiler support
Fixed
- Fixes some ctest failures
Port to Fujitsu and a few bug fixes
Added
- Fujitsu compiler support
Fixed
- This fixes a small CMake bug which can lead to posix_predefined.x being built in the wrong build subdirectory when CMAKE_RUNTIME_OUTPUT_DIRECTORY is set*.
- Missing implementation of
assertIsFinite_real80()
. Apparently undetected until recent attempt to port to flang. - Made support for REAL128 optional. (Port to nvfortran)
Changed
- Updated the CI to use Intel LLVM compilers
- Removed obsolete documentation
Update submodule for fArgParse
Changed
- Updated submodule for fArgparse [v1.6.0]
Workarounds for gfortran 13.2
Fixed
- Several workarounds added to enable building with gfortran 13.2. Polymorphic assignment is broken, and must be replaced by
ALLOCATE(obj,source=...)
. But apparently not everywhere? - Add
-check nouninit
for Intel LLVM to work aroundifx
bug.
Changed
- Updated CI to remove gcc-9 from macOS11 and add gcc-12
Minor bug fixes
Fixed
- Fixed cmake issue where target "pfunit-mpi-defines" is defined more than once
- Missing variable declaration in parameterized test case boiler plate code.
Minor bug fix
Fixed
- Restored consistent behaviour for file paths between
add_pfunit_ctest
amdadd_pfunit_sources
.
Now theadd_pfunit_ctest
handles relative filepaths (e.g../path/to/source.pf
) as described
in the script documentation.
Minor bug fix
Fixed
- Increased size of buffer for reporting real values in asserts. Previous length was not quite enough for some 128 bit values, which resulted in EOR failures during execution.