Internal sign encoding in DISPLAY and PACKED-DECIMAL usage

Hello,

I have a question about internal sign encoding.

"IBM Enterprise COBOL for z/OS - Language Reference - V3R3" says (p.217):

"PACKED-DECIMAL [...] The sign representation uses the same bit
configuration as the 4-bit sign representation in zoned decimal fields."

But "IBM Websphere Studio COBOL for Windows V5.11" gives examples (p.37)
which shows that the 4-bit sign representation depends on usage mode:

My understanding is that:

- For (not separate) DISPLAY usage, it depends on the code:
EBCDIC: + is C, - is D and unsigned is F
ASCII: + and unsigned is 3, and - is 7.

- For PACKED-DECIMAL usage, it is fixed: + is C, - is D and unsigned is F.

Am I right ?

Thanks

--
Vincent Danion
Please, forgive my english !

Vincent,
You have mixed "apples and oranges" - although you may well be correct that
there is an error in the documentation.

What the Enterprise COBOL LRM states *is* true for the Enterprise COBOL
compiler.

It is *NOT* (always) true for IBM's COBOL for Windows.  The sign-nibble on t
he
PC (with the IBM compiler - and several others) is the same between
Packed-Decimal and Zoned Decimal when the compiler is in "EBCDIC" mode but I
S
not the same when in ASCII mode.

Some compilers (for example Micro Focus) has an option for working in ASCII
mode - but using EBCDIC sign-nibbles (or vice versa - I can't remember off t
he
top of my head).  However, this is NOT true for the IBM Windows compiler.

I don't see a separate web page for the "IBM Websphere Studio COBOL for Wind
ows
V5.11" manuals, but the VisualAge COBOL (predecessor product) is at:
http://www-306.ibm.com/software/awd...bol/va/library/

and that seems to indicate that it uses the same LRM as the "mainframe"  (ol
der
version IBM COBOL for OS/390 & VM) LRM.  If that is still the case with the
current Windows compiler, then what you have found is a "bug".

--
Bill Klein
wmklein <at> ix.netcom.com
"Vincent Danion" <vdaNOniSPAMon@scort.com> wrote in message
news:2vuenqF2p8il6U1@uni-berlin.de...
> Hello,
>
> I have a question about internal sign encoding.
>
> "IBM Enterprise COBOL for z/OS - Language Reference - V3R3" says (p.217):
>
> "PACKED-DECIMAL [...] The sign representation uses the same bit configuration
> as the 4-bit sign representation in zoned decimal fields."
>
> But "IBM Websphere Studio COBOL for Windows V5.11" gives examples (p.37) w
hich
> shows that the 4-bit sign representation depends on usage mode:
>
> My understanding is that:
>
> - For (not separate) DISPLAY usage, it depends on the code:
>   EBCDIC: + is C, - is D and unsigned is F
>   ASCII: + and unsigned is 3, and - is 7.
>
> - For PACKED-DECIMAL usage, it is fixed: + is C, - is D and unsigned is F.
>
> Am I right ?
>
> Thanks
>
> --
> Vincent Danion
> Please, forgive my english !

William M. Klein wrote:

Hello,

> NOTE WELL:
>    with current (currently supported) IBM mainframe COBOLs whether or not 
A, B,
> and E are (or are not) considered "valid" sign-nibbles is determined by an
> installation (not user) compiler option.

As far as I have understood, it seems that most compilers are accepting
different nibbles as valid, but use precise values when they are coding
the numbers.

> See:
>   http://publibz.boulder.ibm.com/cgi-...S/IGY3C120/2.41

Thanks. Cobol compilers seems to be highly configurable ! Do you see any
reason (interoperability ?) why people would have configured a specific
sign encoding ?

For instance, OS/400 compilers generally use F for unsigned or
positives, except if CHGPOSSGN option is set (in this case, they use C).
Does anybody know if this option is often used on iSeries ?

> This, of course, has nothing to do with what Assembler or machine instruct
ions
> allow.  There are also rules about how other compiler options impact sign
> handling for COBOL - and there used to be a "usermod" for older COBOL's to
 do
> similar things.

Yes, I understand.

Regarding my initial question, we have done the test (under VA Cobol for
a CICS on NT): the sign nibble is trailing as expected. Therefore, there
is a bug in "IBM Websphere Studio COBOL PGM for Windows V5.11" p.37:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
71 32 33 34
</quote>

should be read as:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
31 32 33 74
</quote>

Have a nice w

--
Vincent Danion.

William M. Klein wrote:

Hello,

> NOTE WELL:
>    with current (currently supported) IBM mainframe COBOLs whether or not 
A, B,
> and E are (or are not) considered "valid" sign-nibbles is determined by an
> installation (not user) compiler option.

As far as I have understood, it seems that most compilers are accepting
different nibbles as valid, but use precise values when they are coding
the numbers.

> See:
>   http://publibz.boulder.ibm.com/cgi-...S/IGY3C120/2.41

Thanks. Cobol compilers seems to be highly configurable ! Do you see any
reason (interoperability ?) why people would have configured a specific
sign encoding ?

For instance, OS/400 compilers generally use F for unsigned or
positives, except if CHGPOSSGN option is set (in this case, they use C).
Does anybody know if this option is often used on iSeries ?

> This, of course, has nothing to do with what Assembler or machine instruct
ions
> allow.  There are also rules about how other compiler options impact sign
> handling for COBOL - and there used to be a "usermod" for older COBOL's to
 do
> similar things.

Yes, I understand.

Regarding my initial question, we have done the test (under VA Cobol for
a CICS on NT): the sign nibble is trailing as expected. Therefore, there
is a bug in "IBM Websphere Studio COBOL PGM for Windows V5.11" p.37:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
71 32 33 34
</quote>

should be read as:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
31 32 33 74
</quote>

Have a nice w

--
Vincent Danion.

William M. Klein wrote:

Hello,

> NOTE WELL:
>    with current (currently supported) IBM mainframe COBOLs whether or not 
A, B,
> and E are (or are not) considered "valid" sign-nibbles is determined by an
> installation (not user) compiler option.

As far as I have understood, it seems that most compilers are accepting
different nibbles as valid, but use precise values when they are coding
the numbers.

> See:
>   http://publibz.boulder.ibm.com/cgi-...S/IGY3C120/2.41

Thanks. Cobol compilers seems to be highly configurable ! Do you see any
reason (interoperability ?) why people would have configured a specific
sign encoding ?

For instance, OS/400 compilers generally use F for unsigned or
positives, except if CHGPOSSGN option is set (in this case, they use C).
Does anybody know if this option is often used on iSeries ?

> This, of course, has nothing to do with what Assembler or machine instruct
ions
> allow.  There are also rules about how other compiler options impact sign
> handling for COBOL - and there used to be a "usermod" for older COBOL's to
 do
> similar things.

Yes, I understand.

Regarding my initial question, we have done the test (under VA Cobol for
a CICS on NT): the sign nibble is trailing as expected. Therefore, there
is a bug in "IBM Websphere Studio COBOL PGM for Windows V5.11" p.37:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
71 32 33 34
</quote>

should be read as:

<quote>
For "PIC S9999 DISPLAY", the value -1234 is encoded like this (NATIVE):
31 32 33 74
</quote>

Have a nice w

--
Vincent Danion.