Merged mjsousa/matiec into default

fix bug introduced by commit 91bef6704b44 (parsing SFCs with transitions in ST and leading whitespace => endless loop)

Only allow function invocation outside ST expressions when command line option is activated.

Ignore pragmas inside IL code when handling inline function calls.

Fix bug: correctly handle formal function calls in IL with no parameters.

Add support for functions returning VOID (i.e. non-standard extension, allowing functions that do not return any data)

Add -warn flag when running flex -> warn of potential bugs

Add support for non-standard VOID datatype. May ony be used to define functions returning VOID.

Delete the YY_NO_UNPUT define, as we do use the unput() function.

Add support for non standard feature: allow POUs with no in, out and inout parameters

Add option to not add the EN and ENO parameters to every function and FB (generated C code becomes smaller, and uses less memory for function invocations)

Correctly identify errors when parsing erroneous code (make sure flex goes back to INITIAL state when code contains errors that do not allow determining whether ST or IL is being parsed)

Created new branch fix_compilation_warnings

fix compilation error about missing braces

IEC_TIMESPEC has two fields tv_sec and tv_nsec.

Compilation error:
./build/POUS.c:600:45: error: missing braces around initializer [-Werror=missing-braces]
static const STEP temp_step = {{0, 0}, 0, {0, 0}};
^
compilation terminated due to -Wfatal-errors.

fix warning about overflow if matiec is running on 64-bit platform,
but the generated C code will cross-compiled for 32-bit platform

For example:
./build/config.c:29:39: error: large integer implicitly truncated to unsigned type [-Werror=overflow]
unsigned long greatest_tick_count__ = 18446744073709551611UL; /*tick*/
^
compilation terminated due to -Wfatal-errors.
cc1: all warnings being treated as errors

fix C compilation warning about function prototype in __SET_EXTERNAL

Example warning:
./POUS.c:1220:3: error: function declaration isn't a prototype [-Werror=strict-prototypes]
__SET_EXTERNAL(data__->,KN_AUT,,__GET_VAR(data__->KEY_AUT.STATE_OUT));
^
compilation terminated due to -Wfatal-errors.
cc1: all warnings being treated as errors

fix C compilation warning from gcc about /* inside of comment section.


./beremiz/iec_std_lib.h:240:1: error: "/*" within comment [-Werror=comment]
/*
^
compilation terminated due to -Wfatal-errors.
cc1: all warnings being treated as errors

Update README.build file with instructions for crosscompiling to win32

Delete commented out code.

Fix FB calls in IL (FB variable name was showing up extraneously before the C function that initializes the POU containing the FB call)

Delete unused code.

Add check for repeated elements in a CASE statement. Emit warnings (and not errors) if found.

transform CASE into if()..else if().. -> fixes 2 bugs (1)case within case statements, and (2)repeated elements in case options

Initialise arrays to their correct default values - Fix bug introduced in commit 477393b00f95 (2014-11-16)

Fixed reset of current_varqualifier when generating global initializers, causin RETAIN and other option to be ORed and applied to all subsequent global variables.

Do fill/narrow datatype analysis algorithm for derived_datatype_identifier_c (fixes bug related to arrays of arrays that was generating incorrect C code).

Better identification of error location in some error messages.

Fix bug introduced in 2014/10/19, while adding support for REF() opertors -> datatype checking was not catching datatype inconsistency errors!

merge

Re-write and fix the logic of SFC action modifiers.

Correct the number given to actions in the #define's

Do not allow constant_propagation algorithm go into infinite loop when analysing code with circular references.

Do constant propagation of FBs instantiated in other FB or Programs, so that variable sized arrays can be declared in FBs too.

Do constant propagation of instantiated FBs, so that variable sized arrays can be declared in FBs too.

Let get_datatype_info_c handle variable sized arrays, even if constant folding is not yet done.

Correctly analyse external<->global var compatibility when using variable sized arrays (TODO: does not yet work for FBs!)

Add comment about bug that needs to be fixed later on.

Break constant_folding_c in two classes: constant_folding_c and constant_propagation_c

Add check whether the same constant var_external variable is initialised with two (or more) different constant values.

Change algorithm for propagating VAR_GLOBAL constant values to corresponding VAR_EXTERNAL.

Disable constant propagation again (still not working correctly)

Make support for use of variables in array size declarations a command line option (off by default)

Add debug_c::print(const_value_c) for printing of const values.

Add is_undefined() method to const_value_c

Make error message more informative.

Add comments.

Add checks for use of non constant value in subranges, and lower_limit > upper_limit.

Fix previous commit (symtable operator[] must return a reference to the stored value!)

Do constant propagation of configuration/resource variables, taking into account scope of variables.

Change dsymbtable_c -> use design pattern used by C++ standard library (STL)

Change symbtable_c -> use design pattern used by C++ standard library (STL)

Remove debugging code left in by mistake.

Do constant propagation inside Configurations and Resources (required for support of extension: array [1..max] OF int); (Resources is still buggy)

Do constant folding of default initial values of datatypes.

Do constant propagation to symbolic_variables, and correctly handle constant folding and propagation of variable declarations.

When generating C code, use the const_value of a symbolic_variable when it is used in a subrange.

Small code re-organization of how const values are stored in symbol_c (can now be accessed through member functions, instead of macros)

Fix bugs introduced in previous commit (once again do constant folding of all literals in type declarations)

Do constant folding of variable's initial value (allows correct C code generation with variables in the subrange of an array declaration: ARRAY [1..max] of INT).

Delete un-used declaration.

merge

Allow variables to be used when declaring the dimension of an array.

Fix C code generation when forward references (-p command line option) is on.

Add a new pou_typename_c object to the AST to store references to previously declared Functions, FB, and Programs.

Delete unused file.

Make use of pre-parser a command line option.

Fix AST generated by parsing with preparsing switched OFF (typo in array datatypes, and fix parsing of POUs).

Fix AST generated by parsing with preparsing switched on, and allow parsing to work with preparsing off.

Fix bug introduced when adding support for relaxed datatype model: change derived_datatype_identifier_c to identifier_c when used as an identifier

Add a pre-parsing phase to stage1_2 (allows source code that references POUs and datatypes before they are declared)

Comment out unused and buggy code.

Update the comments in source code.

Fix bug introduced when adding support for relaxed datatype model. Code cleanup.

Fix main state machine in flex (states were being pushed without being poped from the stack)

merge

Make the relaxed datatype model a runtime option (off by default)

Add support for relaxed datatype model for array datatypes.

fix bug introduced a few commits ago: do not print the first char ('%') of a direct_variable_c (e.g. skip % in %IW4.2)

fix code generation when dereferencing pointers (ref_to) to arrays (ref_to_array^[5])

Fix bug: set the scope annotation in deref_operators used inside structs

Editing of error message to make it more clear.

Filling of symbol->scope annotation moved from narrow to fill_candidate_datatypes_c (became possible because narrowing of struct, array and symbolic variables is now done directly in fill_candidate_datatypes_c)

fill_candidate_datatypes_c now uses search_varinstance_decl_c instead of search_varfb_instance_type_c (moving towards deprecation of search_varfb_instance_type_c)

Do the full fill_candidate/narrow_datatype algorithm for all fields of a structured variable in the fill_candidate_datatypes_c
Allows us to remove temporary code from the search_varfb_decl_c

Fix bug with multple dereferencing (bool_var := bool_ptr_ptr^^;)

Fix bug: Add support for de-referencing pointers to struct (struct_ptr^.elem1) when used inside FB.
Note that multiple de-referencing of structs (struct_ptr_ptr^^.elem) is not supported inside FB code (this would need BIG changes to the compiler!)

Fix bug: add support for dereferencing of REF_TO to REF_TO xxx (e.g.: bool_var := ref_to_ref_to_bool^^;)

Turn support of derefencing operator '^' an option (default is not supported).

Add support for de-referencing of REF_TO datatypes (Note: dereferencing of arrays and structs not yet supported!)

Add support for REF_TO inside arrays and structs (a non-standard feature!)

Code cleanup: with the introduction of anotation_map[], generate_c_vardecl_c can now derive directly from generate_c_base_c

Fix typos in comments

Add comments, and make get_datatype_id_str_c equivalent to get_datatype_id_c

Fix calling of functions whose parameters are of an implicitly declared datatype (currently only makes sense for REF_TO datatypes, but may make sense to other datatypes too if the datatype model is changed in the future).

fix indentations.

in lvalue, add check for REF() fucntion invocation, and the NULL literal

Initialise REF_TO variables to NULL.

Make support for REF_TO ANY a command line option.

Add support for non standard REF_TO ANY (equivalent to void *)

Update some comments. (No changes to code)

Introduce absyntax class for keyword ANY, and use REF_TO(ANY) as datatype of NULL literal.

Fix bug (introduced a few commits ago) when generating SFC actions that are mapped onto variables --> Make generate_sfcdecl_c stateless (i.e. does not depend on order by which it is called).

Add support for the 'NULL' keyword, defined in version 3 of IEC 61131-3.

for option '-Op' (generate <pouname>.h/c): add '#include <pouname>.h' to the .c file, and '#ifdef <pouname>_H' to the .h file

Ass include of iec_std_lib.h in POUS.h file.

Add support for implicitly declared REF_TO datatypes.

With -Op option --> Place implicitly defined datatypes in the .h file corresponding to the POU where they are used.

Add name of POU as prefix to names of implicitly defined datatypes

Implicitly defined array datatypes use new naming method (greatly simplifies stage 4 code).

array_specification_c may be used as a datatype! Add it to spec_init_separator_c

Change REF() operator to return the correct REF_TO datatype datatype.

Narrow array subscripts correctly, even in the presence of other datatype errors (so we do not generate error messages for array subscripts that do not contain errors).

Add support for REF_TO derived datatypes (along with datatype verification)

Add option to place each POU in a distinct .c and .h file pair.

code re-organisation. Preparing for the changes to come in the next commit (one C file for each POU)

Remove un-used parameter from class constructor.

Move standard FBs to a pre-converted C file, insted of converting them every time along with the IEC source.

Move library files compiled to C to their own subdiectory.

Fix stupid mistake in previous commit (use 'long double' instead of 'ld' in istandard library macro)

merge

Update build/compile instructions to reflect the fact that autoconf needs to be executed.

Fix time conversion function in std library, and change into macro (so it may be used in variable initialisations)

Add support for the stepname.T syntax.

Change the name of the variable with the step state to stepname.X (in the generated VARIABLES.csv file)

Add support for Stepname.X syntax.

Remove dead code previously left in by mistake.

Add error message for invalid datatype of transition condition. Generalize datatype narrowing algorithm for transition conditions.

merge