[libgfortran,patch] Wide char part 4, library support

[FX]

This patch provides a character(kind=4) version of the functions in  
intrinsics/string_intrinsics.c, as well as a few needed definitions  
in libgfortran.h. gfc_char4_t is the name of the 32-bit character  
type in the library (it's simply a GFC_UINTEGER_4, which we already  
assume to be available on all targets). I added a  
GFC_SIZE_OF_CHAR_KIND macro which expands to the size in bytes of a  
character variable of a given kind (which, for the two types we  
support, actually is the kind itself).

The new functions are suffixed with "_char4" (as "_c4" is already  
used for complex). To avoid breaking libgfortran's ABI, I have not  
renamed the default character kind functions with "_char1", though it  
could be done next time we break compatibility if it's deemed useful.

To avoid code duplication, all the code in intrinsics/ 
string_intrinsics.c has been put into intrinsics/ 
string_intrinsics_inc.c, and the (relatively few) things that depend  
on the character kind have been macroized: the type itself (and the  
corresponding unsigned type), the function suffixes, the function  
that performs memset(); by the way, I added a helper function to  
replace memset() for wide characters: it's called (you guessed)  
memset_char4. Also, various sizes (those in malloc() and memcpy())  
have been multiplied by the size of the character type (sizeof 
(CHARTYPE)).

Apart from that, really not much. Oh, I've changed the code in one  
place where we did:

-      *dest = internal_malloc_size (*len);
-      memmove (*dest, src, *len);

There's no need to memmove() on a freshly allocated address, so I  
changed that to memcpy().

Of course, you now wonder why doing the library before the rest of  
the front-end. Well, it so happens that I actually have done the  
front-end bits too, so I now almost have a feature-complete widechar  
gfortran (except I/O, for which I'll need Jerry's or Janne's help),  
and I could test these intrinsics and they behave fairly well so far.  
But, the front-end patch still has a few bugs, so I submit this  
easier and self-contained patch while I fix the last bugs in the  
front-end.

Bootstrapped and regtested on x86_64-linux, OK to commit?

FX

-- 
François-Xavier Coudert
http://www.homepages.ucl.ac.uk/~uccafco/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: wide_char_part4.ChangeLog
Type: application/octet-stream
Size: 752 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20080509/4da7dedb/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: wide_char_part4.diff
Type: application/octet-stream
Size: 21588 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/fortran/attachments/20080509/4da7dedb/attachment-0001.obj>