[open-cobol-list] CVS Update 20050702

[Roger While]

CVS Updated. (At July 2 21:10 Central European Time)
(Remember Sourceforge CVS takes a couples of hours to update)

Firstly a word of thanks to people who have been
contributing code/suggestions/bugs off-list :
The guys at AT&T -
    David Korn
    Robert Sherry
    Glenn Fowler  (the same as the one of FVN hash fame ?)

Then :
Karst Dearth
Rich Painter

Some of the code/suggestions will be incorporated before the
0.33 release (some is included in this update).
Other parts will be considered/implemented post-0.33.

CVS Changes :

Fix the reported problems from the Bugs list.
Specifically :
The early phrase detection in the scanner was missing
the white space at the end of the phrase(s) leading to
incorrect parsing.
COPY REPLACING was not working correctly.

Fix segmentaion fault or similar when trying to use IS/SORT
when DB is not configured.

Specifically cast offsets in fseek calls to off_t.

Put in a bunch of Intrinsic Function tests in the OC tests ("make check")
(Note, some of these return implicitly COMP-2, aka FLOAT-LONG, aka 
"double",
 so there may be small differences reported in the least significant 
places under
 different environments. If tests fail, check int the run.dir/xx 
directories, where xx
 is the failed test number, for the ".log" file. This details expected and 
actual results)

A few cleanups/fixes in the intrinsic code.

Put in some experimental code in move.c/numeric.c -
At least upto gcc 3.4, "memset" never inlines, no matter what
options are thrown at gcc.
"memcpy" may inline if you throw the right options at gcc.
Define our own simple versions of these. May eventually
be replaced with asm code on Intel and possibly merged with
the byteswap code. The eventaul aim would be to get the
cob_binary_get/set routines totally inlined.

Roger