Syntax error in Frama-C due to custom machdep

Question

I am using MPLAB XC16 C Compiler for my application. If I use machdep x86_16, the Frama-C works normally. For example, I can launche Frama-C in this way:

$ frama-c-gui machdep x86_16 -cpp-command 'C:\\"Program Files (x86)"\\Microchip\\xc16\\v1.26\\bin\\xc16-gcc.exe -E' -no-cpp-gnu-like D:\\project\\*.c

But machdep x86_16 do not comply fully with XC16. So I want to customize machdep.
Following the instructions, I created file machdep_xc16.ml that contain:

open Cil_types

let xc16 =
{
  version          = "dsPIC33F";
  compiler         = "XC16";    (* Compiler being used. *)
  sizeof_short     = 2;         (* Size of "short" *)
  sizeof_int       = 2;         (* Size of "int" *)
  sizeof_long      = 4;         (* Size of "long" *)
  sizeof_longlong  = 8;         (* Size of "long long" *)
  sizeof_ptr       = 2;         (* Size of pointers *)
  sizeof_float     = 4;         (* Size of "float" *)
  sizeof_double    = 4;         (* Size of "double" *)
  sizeof_longdouble  = 8;       (* Size of "long double" *)
  sizeof_void      = 0;         (* Size of "void" *)
  sizeof_fun       = 0;         (* Size of function *)
  size_t = "unsigned int";      (* Type of "sizeof(T)" *)
  wchar_t = "unsigned short";   (* Type of "wchar_t" *)
  ptrdiff_t = "int";            (* Type of "ptrdiff_t" *)
  alignof_short    = 2;         (* Alignment of "short" *)
  alignof_int      = 2;         (* Alignment of "int" *)
  alignof_long     = 2;         (* Alignment of "long" *)
  alignof_longlong = 2;         (* Alignment of "long long" *)
  alignof_ptr      = 2;         (* Alignment of pointers *)
  alignof_float    = 2;         (* Alignment of "float" *)
  alignof_double   = 2;         (* Alignment of "double" *)
  alignof_longdouble = 2;       (* Alignment of "long double" *)
  alignof_str      = 1;         (* Alignment of strings *)
  alignof_fun      = 1;         (* Alignment of function *)
  alignof_aligned  = 16;        (* Alignment of a type with aligned attribute *)
  char_is_unsigned = false;     (* Whether "char" is unsigned *)
  const_string_literals = true; (* Whether string literals have const chars *)
  little_endian = true;         (* whether the machine is little endian *)
  underscore_name = true;       (* If assembly names have leading underscore *)
  has__builtin_va_list = false; (* Whether [__builtin_va_list] is a known type *)
  __thread_is_keyword = false;  (* Whether [__thread] is a keyword *)
}

let mach2 = { xc16 with compiler = "baz" }

let () =
  let ran = ref false in
  Cmdline.run_after_loading_stage
    (fun () ->
       Kernel.result "Registering machdep 'xc16' as 'XC16'";
       File.new_machdep "XC16" xc16;
       if !ran then begin
         Kernel.result "Trying to register machdep 'mach2' as 'XC16'";
         File.new_machdep "XC16" mach2
       end
       else ran := true
    )

I inserted the following lines in the file __fc_machdep.h just before line "#error Must define ..."

#ifdef __FC_MACHDEP_XC16
    #define  __FC_BYTE_ORDER __LITTLE_ENDIAN

    /* min and max values as specified in limits.h */
    #define __FC_SCHAR_MAX  0x7f
    #define __FC_SCHAR_MIN  (-__FC_SCHAR_MAX -1)
    #define __FC_UCHAR_MAX  0xff
    #define __FC_CHAR_MIN   __FC_SCHAR_MIN
    #define __FC_CHAR_MAX   __FC_SCHAR_MAX

    #define __FC_SHRT_MAX   0x7fff
    #define __FC_SHRT_MIN   (-__FC_SHRT_MAX -1)
    #define __FC_USHRT_MAX  0xffff
    #define __FC_INT_MAX    __FC_SHRT_MAX
    #define __FC_INT_MIN    __FC_SHRT_MIN
    #define __FC_UINT_MAX   __FC_USHRT_MAX

    #define __FC_LONG_MAX   0x7fffffff
    #define __FC_LONG_MIN   (-__FC_LONG_MAX -1)
    #define __FC_ULONG_MAX  0xffffffffU

    #define __FC_LLONG_MAX  0x7fffffffffffffffLL
    #define __FC_LLONG_MIN  (-__FC_LLONG_MAX -1)
    #define __FC_ULLONG_MAX 0xffffffffffffffffUL

    /* Required */
    #undef  __CHAR_UNSIGNED__
    #define __WORDSIZE 16                   
    #define __SIZEOF_SHORT 2
    #define __SIZEOF_INT 2
    #define __SIZEOF_LONG 4
    #define __SIZEOF_LONGLONG 8
    #define __CHAR_BIT 8
    #define __PTRDIFF_T int
    #define __SIZE_T unsigned int
    #define __FC_SIZE_MAX __FC_INT_MAX

    /* stdio.h */
    #define __FC_EOF (-1)
    #define __FC_FOPEN_MAX 8
    #define __FC_RAND_MAX 32767
    #define __FC_PATH_MAX 260
    #define __WCHAR_T unsigned short

    /* Optional */
    #define __INT8_T signed char
    #define __UINT8_T unsigned char
    #define __INT16_T signed int
    #define __UINT16_T unsigned int

    #define __INTPTR_T signed int
    #define __UINTPTR_T unsigned int
    #define __INT32_T signed long
    #define __UINT32_T unsigned long
    #define __INT64_T signed long long
    #define __UINT64_T unsigned long long

    /* Required */
    #define __INT_LEAST8_T signed char
    #define __UINT_LEAST8_T unsigned char
    #define __INT_LEAST16_T signed int
    #define __UINT_LEAST16_T unsigned int
    #define __INT_LEAST32_T signed long
    #define __UINT_LEAST32_T unsigned long
    #define __INT_LEAST64_T signed long long
    #define __UINT_LEAST64_T unsigned long long

    #define __INT_FAST8_T signed char
    #define __UINT_FAST8_T unsigned char
    #define __INT_FAST16_T signed int
    #define __UINT_FAST16_T unsigned int
    #define __INT_FAST32_T signed long
    #define __UINT_FAST32_T unsigned long
    #define __INT_FAST64_T signed long long
    #define __UINT_FAST64_T unsigned long long

    /* POSIX */
    #define __SSIZE_T signed long

    #define __FC_PTRDIFF_MIN __FC_INT_MIN 
    #define __FC_PTRDIFF_MAX __FC_INT_MAX
    #define __FC_VA_LIST_T char*

    /* Required */
    #define __INT_MAX_T signed long long
    #define __UINT_MAX_T unsigned long long
#else

Now if I launch Frama-C in this way:

$ frama-c-gui -load-script machdep_xc16 -machdep XC16 -cpp-command 'C:\\"Program Files (x86)"\\Microchip\\xc16\\v1.26\\bin\\xc16-gcc.exe -E' -no-cpp-gnu-like D:\\project\\*.c

I get output like this:

[kernel] Registering machdep 'xc16' as 'XC16'
[kernel] Parsing .opam/4.02.3+mingw64c/share/frama-c/libc/__fc_builtin_for_normalization.i (no preprocessing)
[kernel] warning: machdep XC16 has no registered macro. Using __FC_MACHDEP_XC16 for pre-processing
[kernel] Parsing D:/project/main.c (with preprocessing)
. . .
[kernel] Parsing D:/project/get_data.c (with preprocessing)
[kernel] syntax error at .opam/4.02.3+mingw64c/share/frama-c/libc/__fc_define_wchar_t.h:28:
         26    #if !defined(__cplusplus)
         27    /* wchar_t is a keyword in C++ and shall not be a typedef. */
         28    typedef __WCHAR_T wchar_t;
               ^^^^^^^^^^^^^^^^^^^^^^^^^^
         29    #else
         30    typedef __WCHAR_T fc_wchar_t;

The syntax error occurs when the file containing #include <stdio.h> is processed.
What am I doing wrong?

Answer 1

The instructions about how to add a new machdep have been revised in the manual and will be available on the next Frama-C release (Phosporus).

The main issue with a new machdep is that there are two (seemingly redundant) parts to a machdep: the OCaml-level definitions, used by Frama-C, and the C-level definitions, used by the C preprocessor while parsing the Frama-C standard library. Realizing that both are necesssary and complementary helps understanding why the whole process is cumbersome (although it will be simplified in the future).

Here's an extract of the upcoming instructions:

A custom machine description may be implemented as follows:

let my_machine = {
  version          = "generic C compiler for my machine";
  compiler         = "generic";  (* may also be "gcc" or "msvc" *)
  cpp_arch_flags   = ["-m64"];
  sizeof_short     = 2;
  sizeof_int       = 4;
  sizeof_long      = 8;
  (* ... *)
}

let () = File.new_machdep "my_machine" my_machine

Note that your machdep_xc16.ml can be simplified: the code you used is part of a test that tries to register twice the same machdep, just to ensure that it fails. But in practice, when you use -load-script you can just create the machdep directly as above, calling File.new_machdep directly.

After this code is loaded, Frama-C can be instructed to use the new machine model using the -machdep command line option.

If you intend to use Frama-C's standard library headers, you must also do the following:

• define constant __FC_MACHDEP_<CUSTOM>, replacing <CUSTOM> with the name (in uppercase letters) of your created machdep; this can be done via -cpp-extra-args="-D__FC_MACHDEP_<CUSTOM>";

• provide a header file with macro definitions corresponding to your caml definitions. For the most part, these are macros prefixed by __FC_, corresponding to standard C macro definitions, e.g., __FC_UCHAR_MAX. These definitions are used by Frama-C's <limits.h> and other headers to provide the standard C definitions. The test file tests/misc/custom_machdep/__fc_machdep_custom.h contains a complete example of the required definitions. Other examples can be found in share/libc/__fc_machdep.h.

Make sure that your custom header defines the __FC_MACHDEP include guard, and that the program you are analyzing includes this header before all other headers. One way to ensure this without having to modify any source files is to use an option such as -include in GCC.

An example of the complete command-line is presented below, for a custom machdep called myarch, defined in file my_machdep.ml and with stdlib constants defined in machdep_myarch.h:

frama-c -load-script my_machdep.ml -machdep myarch \
   -cpp-extra-args="-D__FC_MACHDEP_MYARCH -include machdep_myarch.h"

Note that the __fc_machdep_custom.h in Silicon is incomplete, but the version you posted seems complete, so use it instead: put it in a file called e.g. machdep_xc16.h, add #define __FC_MACHDEP to it, and include it before the other files, e.g. using -include machdep_xc16.h as preprocessor flag. This will ensure that your version of the machdep will be used instead of Frama-C's, which will then allow you to use Frama-C's standard library with the constants defined according to your architecture.

Also, because your command line contains -cpp-command and -no-cpp-gnu-like, you'll have to adapt the -cpp-extra-args above, putting -D__FC_MACHDEP_MYARCH and -include machdep_myarch.h directly in your -cpp-command.