When defining real to always be double with
DBLFLAG= -fdefault-real-8 -fdefault-double-8
and compiling with llvm flang v21+ (or flang-new in older versions), gFTL cannot find the right stringvector and throws many errors like:
error: Cannot use module file for module 'gftl_stringvector': File is not the right module file for 'gftl_stringvector': $PFUNIT_INSTALL_PATH/GFTL_SHARED-1.12/include/v1/gftl_stringvector.mod
use gftl_stringvector,only:stringvectoriterator
....
$PFUNIT_INSTALL_PATH/FARGPARSE-1.11/include/fp_cast.mod:66:5: error: Cannot parse module file for module 'gftl2_stringvector': Source file 'gftl2_stringvector.mod' was not found
use gftl2_stringvector,only:stringvector
This is fixed by including also V2 of gFTL in cmake/PFUNIT.mk.in (which seems to manage it's own real definition and not depend on the project's definition):
PFUNIT_EXTRA_FFLAGS := -I$(GFTL_INCLUDE_DIR) -I$(GFTL_V2_INCLUDE_DIR) -I$(GFTL_SHARED_INCLUDE_DIR) -I$(GFTL_SHARED_V2_INCLUDE_DIR) -I$(FARGPARSE_INCLUDE_DIR) -I$(PFUNIT_INCLUDE_DIR) -I.
I also tried adding V2 directly into the project's Makefile on top of the PFUNIT_EXTRA_FFLAGS but that didn't seem to work. I have to add them to PFUNIT.mk
I'm not sure if this is a proper fix, but as gFTL v2 is already built by default, is there any reason to not include it by default?
When defining real to always be double with
and compiling with llvm flang v21+ (or flang-new in older versions), gFTL cannot find the right stringvector and throws many errors like:
This is fixed by including also V2 of gFTL in cmake/PFUNIT.mk.in (which seems to manage it's own real definition and not depend on the project's definition):
I also tried adding V2 directly into the project's Makefile on top of the
PFUNIT_EXTRA_FFLAGSbut that didn't seem to work. I have to add them to PFUNIT.mkI'm not sure if this is a proper fix, but as gFTL v2 is already built by default, is there any reason to not include it by default?