Recent changes to bugs

DLList failure in "pop" under IFort and Cray

Michael List — Fri, 30 Sep 2011 02:55:39 -0000

The Intel and the Cray compilers seem displeased with the unit test for "pop" with DLList<integer>. This is not present with GFortran.

Missing template parameters

David Car — Tue, 15 Jun 2010 15:05:03 -0000

If a template parameter is forgotten, PyF95++ does not catch this error and the parsing is not correct. Need to flag the user if a parameter is missing and stop.

Prepod foreach overwritten by multiple occurences

Michael List — Sat, 09 Jan 2010 17:37:13 -0000

Multiple occurrences of foreach results in only the final one being instantiated (but in multiple places...). This is similar to when the occurrence is not given a unique name.

Build Dependency Graph

David Car — Mon, 07 Dec 2009 04:51:16 -0000

PyF95++ does not properly build the dependency graph for a templated module with two different template parameter lists in the same file. For instance, creating two templated modules with derived type Foo; one takes only one parameter and the other has two template parameters. With both template modules in the same file, instantiating and instance of both templated derived types results an error when building the file dependency graph. A work around for the moment is to have each in their own separate file.

Memory leak in tolower and toupper string methods

Michael List — Sun, 06 Dec 2009 17:20:43 -0000

The string methods exhibit a memory leak: toupper and tolower.

Unit Testing only supports one run

Michael List — Fri, 04 Dec 2009 00:11:40 -0000

This looks similar to an issue I hit when I did not put an id into the name of the enumeration block. Basically, only the newest occurrence survives as the old ones are overwritten.

The solution is to use str(id(self)) as a portion of the name. However, this interrupts whatever searches are called when the run template is instantiated and it finds all of the TestRunner blocks, causing it to throw an exception.

Pyparsing grammar does not require 1st occurances

Michael List — Fri, 04 Dec 2009 00:08:29 -0000

The regular expressions that were used specified that only spaces and tabs could be found before a line where a block was defined. With pyparsing this does not exist. This should be corrected by specifying KEYWORD as having a required set of spaces/tabs only prior to a block's invocation.

Example:

! We like enumeration because it's cool

>>>>

! We like
integer, parameter :: because=1
integer, parameter :: it's=2
integer, parameter :: cool=3

NUM does not include negative numbers

Anonymous — Thu, 03 Dec 2009 16:37:01 -0000

The numeric argument parsing for the blocks in F95 does not support negative numbers. It also does not support the use of engineering format numbers.

-1
-1.0e+1
-.001e-9

etc.

F90T files without dependencies in the STL are not written

Michael List — Thu, 03 Dec 2009 01:01:49 -0000

Nominally, F90T files which do not contain templates, usage of STL classes, etc. are not written to F90 files. This makes it difficult to use advanced Fortran features such as enumerations.

This is related to the feature request to mitigate the parsing of the STL.

Closure of blocks

Michael List — Wed, 02 Dec 2009 01:22:41 -0000

When a labeled block is not terminated with a matching label the parse misses the end of the block and thus will not correctly parse. Standard Fortran does not require the matching label (though it is much to our liking) and as a result this should throw at least a warning. Currently the requirement is to label-terminate.

Example:
type Option
...
end type

will not work. Dependencies will not be generated and the type will not terminate in BlockIt. The following is correct:

type Option
...
end type Option