[open-cobol-list] Re: Update from AT&T

[Roger While]

Shoot me a tarball first.
Are the changes based on current CVS ?
Some of this clashes with forthcoming CVS update this weekend.

Comments :
1 a,b,c,d  - OK. cob_get/put_sign will need checking. get changes field 
content
                     for further calculation.
                     Where we really need optimization is with binary 
fields. I have something
                      on the back-burner for this.

2 a/b - I do not understand this. The prototypes in the standard
include files should take care of this. Have you got a borked /usr/include 
?
2 c  - Yes, I agree we should do this.
2 d  - NO, The programmer should know what he/she is doing.
           (And anyway, how do you compatibly redefine a pointer as binary 
?
              32-bitters - 4 bytes, 64-bitters 8 bytes)

3 a  - Maybe use existing relaxed-syntax-check ?
3 b  - Ditto, and needs checking; if allowed, code as it stands generates
                     wrong storage-allocation.
3 c  - OK
3 d  - I think this may give wrong results in certain cases. Needs 
checking.

4 a  - Implementation (partial) coming in CVS update this weekend.
          Table of FUNCTIONS in reserved.c. structs, defs, enums, 
variables in
            tree.h, tree.c. Simple functions implemented in typeck.c.
          In libcob, intrinsic.h, intrinsic.c added.
          FUNCTIONS - E, PI, WHEN-COMPILED, CURRENT-DATE, CHAR,
          LENGTH, DATE-OF-INTEGER. (Maybe some more if I have time before 
the commit)
4 b  - Already been done.

5 a  - Sounds good.
5 b  - Ditto
5 c  - What is this ??
5 d  - If I understand correctly, this is not good enough. You need some 
generated constant
          that can be checked with a libcob function.
5 e  - Actually, the place to do this is in cobcrun.c. Throw in your own 
code, link it with whatever DB/ own
           C routines, put the entry points into call.def (so that a 
normal C function call is generated) and
          away you go. Compile your complete app(s) as modules and you are 
ready.

6 a  -  Things around this have been changed in CVS. configure.ac/configure 
etc.
6 b  - Naaa, Don't you know with what "-std=" the app was compiled ?  :-)
          Anyway, it is quite valid to call between applications that have 
been compiled
          differently. I have apps that do this. (MF app <-> MVS app).
6 c - Do not understand. Do you mean you replaced fp with f->file ?
         If so, this definitely does not produce better code on Intel/gcc.
          Part 2 - Yes, agreed.

7 a/b  -  Presumably you will be maintaining this. How do you deal with 
changes to
              configure(.ac) ?


Roger