Fix OpenMP

• Rewrite of OpenMP recognition in assembler code, also see #17
• Fixed help message for -a all option

Fix release

• Support INTERFACEs within Subroutine definitions
• Bug fixes
• More tests

A little bit of Fortran77

• Support for some Fortran77 syntax
• Prevent tracking of a global variable that is shadowed by a local variable
• Bug fixes
• More tests

Semicolons

• Support for the semicolon syntax. Attention! I have not changed the fact that the LHS of an assignment will be tracked starting from the line following the assignment. That means, in the following statements the use of a%c1 will be missed by the analysis. Needs to be fixed in the future (#13).

p => a ; b = p%c1

• Added more default filename patterns: Module files will now be found automatically when they follow one of the following patterns (assembler files accordingly with .s ending):

moduleName + '.f90'                                              
moduleName + '.F90'                                              
moduleName + '_mod.f90'                                                   
moduleName + '_mod.F90'                                  
moduleName.replace('_mod', '') + '.f90'                                                   
moduleName.replace('_mod', '') + '.F90'

• Bug fixes

• More tests

• Refactorings

Full Types

Always Full Types

Introduced new config variable ALWAYS_FULL_TYPES. There, one can choose some types for which always all components shall be considered as "used".

Example

Assume you have a the following type:

TYPE :: my_type
  INTEGER :: i1, i2
END my_type

TYPE :: other_type
  REAL :: r
END my_type

and in your FCG configuration:

ALWAYS_FULL_TYPES = {'my_type', 'other_type'}

For every argument of type my_type the components %i1 and %i2 are considered to be used in the analyzed subroutine without any further tracking, and accordingly for %r of other_type. When there is global variable or a component of another argument or global variable with type my_type (or other_type) and this global variable or component appears somewhere in the subroutine, the (sub-)components %i1 and %i2 (or %r) are considered to be used without any further tracking.

---

Plus

• Bug fixes
• More tests
• Refactorings

Abstract Types

Handling of Abstract Types

Introduced new config variable ABSTRACT_TYPE_IMPLEMENTATIONS. There, one can choose for some abstract types the one and only implementing subtypes.

Example

ABSTRACT_TYPE_IMPLEMENTATIONS = {'my_abstract_type':('my_subtype_module','my_subtype'),
                                 'other_abstract_type':('other_subtype_module','other_subtype'), 
                                 ...}

FCG will than handle every variable with type CLASS(my_abstract_type) as if it were actually TYPE(my_subtype) and every variable with type CLASS(other_abstract_type) as if it were actually TYPE(other_subtype) and so on....

Only helps when my_subtype or other_subtype are actually the chosen one, for example depending on some configuration. When during run time objects of several subtypes of my_abstract_type or other_abstract_type come along, it is not very helpful. (In future, maybe multiple subtypes will be supported, including automatic retrieval of all existing subtypes.)

---

Plus

• Bug fixes
• More tests

Python 3

• Runs with Python 3 now (Fixes #3) :-)

• New exit codes for errors

• Subroutines are now sorted by first occurrence when printing call tree

Minor changes

• Bug fixes
• Code clean up
• New test suite

Function Results

• Type components of function results are now analyzed
• Added option -a result
• Enhanced parsing of variable declaration
• Refactorings
• Added tests
• Code clean up
• Fixes #5
• Updated documentation

OpenMP

• Assembler code containing OpenMP regions is now supported.
• Fixed error introduced in latest release regarding the tracking of global variables as function results.
• Removed some deprecated functions.

Fixes #7.