On 02-Nov-2015 17:01 -0600, Jon Paris wrote:
On Nov 2, 2015, at 4:58 PM, CRPence wrote:
Was not that overflow error prevented with the Truncate Numeric
(TRUNCNBR) specification of *YES, so as to "Ignore numeric overflow
and move the truncated value to the result field" [to mimic] just
like what happened in the older RPG? And that parameter
specification defaulted to *YES, so that problem would not become
an issue for that calculation with the newer compiler, unless
explicitly changed was the default or the explicit specification to
the value of *NO, asking to cause the overflow error to be exposed
at run-time? I was too lazy to test; but interested enough to
overcome my laziness to look up the parameter name, the parameter
help text, and the default :-)
No Chuck - in fact the exact opposite.
The point of TRUNCNBR is to make RPG III style op-codes such as
MULT/ADD/etc. behave the _same_way_ as their more intelligent RPG IV
operations i.e. to fail in the event of an overflow condition.
There is no Truncate Numeric feature on the OPM compiler command that
I can find. I can find that parameter only on the ILE compiler
commands. Perhaps thinking instead, of the Ignore Decimal Data Errors
(IGNDECERR) parameter?
Just to confirm:
I compiled the following source having SRCTYPE(RPG), with the
14=Compile the command CRTRPGPGM effected a compile with defaults:
FQSQPTABLIP F 4 DISK
IQSQPTABLXX
C Z-ADD110215 MMDDYY 60
C MMDDYY MULT 10000.01 YYMMDD 60
C Z-ADD151102 YMD6 60
C YYMMDD IFEQ YMD6
C 'ALL GOOD'DSPLY
C ENDIF
The effect for CALL of the program was displaying the text "ALL GOOD".
I then converted the above source using CVTRPGSRC, which effected the
following source [member having SRCTYPE(RPGLE)], after which with the
14=Compile the command CRTBNDRPG effected a compile with defaults:
FQSQPTABL IP F 4 DISK
IQSQPTABL XX
C Z-ADD 110215 MMDDYY 6 0
C MMDDYY MULT 10000.01 YYMMDD 6 0
C Z-ADD 151102 YMD6 6 0
C YYMMDD IFEQ YMD6
C 'ALL GOOD' DSPLY
C ENDIF
Just like the pre-converted program, the effect for CALL of the new
program was displaying the text "ALL GOOD". There was no error.
To force the msg RNQ0103 to transpire instead, during run-time in
response to the msg MCH1210, I had to explicitly specify the value *NO
for the Truncate Numeric (TRUNCNBR) on the compile request.
Seems the default behavior for the new compiler was to honor [to
mimic] the old\bad way of ignoring overflow for the simple MULT
operation [while also moving the truncated result to the result field].
That seems only natural, to best protect someone who might be using
the CVTRPGSRC and then just compiling their new source to replace the
old; i.e. their expectation is that the program created [from the
post-cvt src] with the new compiler should work the same as the program
created [from the pre-cvt src] with the old compiler. From the help
text is also the "Note: The TRUNCNBR option does not apply to
calculations performed within expressions. (Expressions are found in the
Extended-Factor 2 field.) If overflow occurs for these calculations, an
error will always occur."; but that would not apply to a
merely-converted source, because the older RPG did not have that
capability, and CVTRPGSRC will not generate any code changes, for
example to make use of an EVAL.
midrange.com
