Merge pull request #97 from GEOS-ESM/develop

Merge develop into master

GNU.cmake

@@ -1,5 +1,5 @@
-if (CMAKE_Fortran_COMPILER_VERSION VERSION_LESS 5.3)
-  message(FATAL_ERROR "${CMAKE_Fortran_COMPILER_ID} version must be at least 5.3!")
+if (CMAKE_Fortran_COMPILER_VERSION VERSION_LESS 8.3)
+  message(FATAL_ERROR "${CMAKE_Fortran_COMPILER_ID} version must be at least 8.3!")
 endif()
 
 set (FOPT0 "-O0")
@@ -22,7 +22,73 @@ set (FREAL8 "-fdefault-real-8 -fdefault-double-8")
 set (FINT8 "-fdefault-integer-8")
 
 set (PP    "-cpp")
+
+# GCC 10 changed behavior in how it handles Fortran. To wit:
+
+#   Mismatches between actual and dummy argument lists in a single file are
+#   now rejected with an error. Use the new option -fallow-argument-mismatch
+#   to turn these errors into warnings; this option is implied with
+#   -std=legacy. -Wargument-mismatch has been removed.
+
+#   The handling of a BOZ literal constant has been reworked to provide
+#   better conformance to the Fortran 2008 and 2018 standards. In these
+#   Fortran standards, a BOZ literal constant is a typeless and kindless
+#   entity. As a part of the rework, documented and undocumented extensions
+#   to the Fortran standard now emit errors during compilation. Some of
+#   these extensions are permitted with the -fallow-invalid-boz, where the
+#   error is degraded to a warning and the code is compiled as with older
+#   gfortran.
+
+# GEOS has many places where, if not converted to warning, the model
+# will not build. Until the code can be fixed, for now we set the flags to
+# convert to warnings. But by using an option() we can allow this to be
+# turned off for testing.
+
+# GCC 9 and lower do not have the flag...
 set (MISMATCH "")
+set (ALLOW_BOZ "")
+
+# With GCC 10...
+if (CMAKE_Fortran_COMPILER_VERSION VERSION_GREATER_EQUAL 10)
+
+   # First for the argument mismatch
+   option(MISMATCH_IS_ERROR "Argument mismatches are errors, not warnings" OFF)
+   if (NOT MISMATCH_IS_ERROR)
+      ecbuild_warn (
+         "Argument mismatches will be treated as *warnings* and not *errors*. "
+         "Per the gfortran 10 man page:\n"
+         "Some code contains calls to external procedures which \n"
+         "mismatches between the calls and the procedure definition, \n"
+         "or with mismatches between different calls.  Such code is \n"
+         "non-conforming, and will usually be flagged wi1th an error. \n"
+         "This options degrades the error to a warning, which can \n"
+         "only be disabled by disabling all warnings vial -w.  Only a \n"
+         "single occurrence per argument is flagged by this warning. \n"
+         "-fallow-argument-mismatch is implied by -std=legacy.\n"
+         "Using this option is *strongly* discouraged.  It is possible to \n"
+         "provide standard-conforming code which allows different types \n"
+         "of arguments by using an explicit interface and TYPE(*).")
+      set (MISMATCH "-fallow-argument-mismatch")
+   endif ()
+
+   # Then for BOZ constants
+   option(INVALID_BOZ_IS_ERROR "Use of invalid BOZ constants are errors, not warnings" OFF)
+   if (NOT INVALID_BOZ_IS_ERROR)
+      ecbuild_warn(
+         "Invalid use of BOZ literal constants will be treated as *warnings* and not as *errors*. "
+         "Per the GCC 10 release notes:\n"
+         "The handling of a BOZ literal constant has been reworked \n"
+         "to provide better conformance to the Fortran 2008 and 2018 \n"
+         "standards. In these Fortran standards, a BOZ literal constant is a \n"
+         "typeless and kindless entity. As a part of the rework, documented \n"
+         "and undocumented extensions to the Fortran standard now emit \n"
+         "errors during compilation. Some of these extensions are permitted \n"
+         "with the -fallow-invalid-boz, where the error is degraded to a \n"
+         "warning and the code is compiled as with older gfortran.")
+      set (ALLOW_BOZ "-fallow-invalid-boz")
+   endif ()
+endif ()
+
 #set (BIG_ENDIAN "-fconvert=swap") # This doesn't seem to work at the moment
 #set (LITTLE_ENDIAN "") # Not sure
 set (EXTENDED_SOURCE "-ffixed-line-length-132")
@@ -50,7 +116,7 @@ add_definitions(-D__GFORTRAN__)
 # Common Fortran Flags
 # --------------------
 set (common_Fortran_flags "-ffree-line-length-none ${NO_RANGE_CHECK} -Wno-missing-include-dirs ${TRACEBACK}")
-set (common_Fortran_fpe_flags "-ffpe-trap=zero,overflow ${TRACEBACK}")
+set (common_Fortran_fpe_flags "-ffpe-trap=zero,overflow ${TRACEBACK} ${MISMATCH} ${ALLOW_BOZ}")
 
 # GEOS Debug
 # ----------
@@ -69,8 +135,9 @@ set (GEOS_Fortran_Release_FPE_Flags "${common_Fortran_fpe_flags}")
 
 # Options per Jerry DeLisle on GCC Fortran List
 #set (GEOS_Fortran_Vect_Flags "${FOPT2} -march=native -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 -mprefer-avx128 -mno-fma")
+#set (GEOS_Fortran_Vect_FPE_Flags "${DEBINFO} ${TRACEBACK}")
 
-# Options per Jerry DeLisle on GCC Fortran List with SVML
+# Options per Jerry DeLisle on GCC Fortran List with SVML (does not seem to help)
 #set (GEOS_Fortran_Vect_Flags "-O2 -march=native -ffast-math -ftree-vectorize -funroll-loops --param max-unroll-times=4 -mprefer-avx128 -mno-fma -mveclibabi=svml")
 #set (GEOS_Fortran_Vect_FPE_Flags "${DEBINFO} ${TRACEBACK}")