HP 3000 Manuals HP 3000 Manuals
Using Debug (continued)
Debugging Trap Errors
[REV BEG]
Invalid Decimal Data in NUMERIC Class Condition.
This example shows how you can force the NUMERIC class condition on
packed decimal data to be true in certain cases where it would otherwise
be false. By specifying I in column 9 of COBRUNTIME, any NUMERIC class
condition on packed decimal data is true in the following cases:
* A signed value in an unsigned PACKED-DECIMAL field.
* An unsigned value in a signed PACKED-DECIMAL field.
* Any invalid sign nibble.
This matches the behavior of HP COBOL II/V programs. Unless you put I in
column 9 of COBRUNTIME, the above conditions cause any NUMERIC class
condition to be false.
The following is an example program that contains invalid packed decimal
data:
WORKING-STORAGE SECTION.
01 DECIMAL-NO-SIGN PIC 9(3) USAGE PACKED-DECIMAL VALUE 123.
01 DECIMAL-SIGN REDEFINES DECIMAL-NO-SIGN
PIC S9(3) USAGE PACKED-DECIMAL.
PROCEDURE DIVISION.
PARA-001.
DISPLAY "DECIMAL-NO-SIGN is ", DECIMAL-NO-SIGN.
DISPLAY "DECIMAL-SIGN is ", DECIMAL-SIGN.
IF DECIMAL-NO-SIGN IS NUMERIC
DISPLAY "DECIMAL-NO-SIGN is NUMERIC."
ELSE DISPLAY "DECIMAL-NO-SIGN is not NUMERIC."
END-IF.
IF DECIMAL-SIGN IS NUMERIC
DISPLAY "DECIMAL-SIGN is NUMERIC."
ELSE DISPLAY "DECIMAL-SIGN is not NUMERIC."
END-IF.
STOP RUN.
When this program is compiled and run in Native Mode with column 9 of
COBRUNTIME set to anything but I, the program displays the following:
DECIMAL-NO-SIGN is 123
DECIMAL-SIGN is +123
DECIMAL-NO-SIGN is NUMERIC.
DECIMAL-SIGN is not NUMERIC. DECIMAL-SIGN contains an invalid sign.
[REV END]
[REV
BEG]
When this program is compiled and run either in Native Mode with column 9
of COBRUNTIME set to I, or in Compatibility Mode, it displays the
following:
DECIMAL-NO-SIGN is 123
DECIMAL-SIGN is +123
DECIMAL-NO-SIGN is NUMERIC.
DECIMAL-SIGN is NUMERIC. Invalid sign still makes NUMERIC test true.
The following diagram shows the contents of the single data item named by
DECIMAL-NO-SIGN and redefined by DECIMAL-SIGN. The values shown are
hexadecimal. Each box in the diagram represents one packed-decimal
digit, which is 4 bits:
-------------------------------------------------------------------------------------------------
- 1 - 2 - 3 - F -
-------------------------------------------------------------------------------------------------
The rightmost position is the sign of the data value. The F in this
position means the value is unsigned. Any valid signed PACKED-DECIMAL
data has either the hexadecimal value C for a positive value, or D for a
negative value, in the rightmost 4 bits. For more information, see the
section "USAGE IS PACKED-DECIMAL or COMPUTATIONAL-3" in Chapter 7 of
the HP COBOL II/XL Reference Manual.[REV END]
[REV BEG]
Traps with COBOL Functions.
In the following example, the parameter for the FACTORIAL function is out
of range.
00001 000100$control post85
00001 001000 identification division.
00002 001100 program-id. trap-fact.
00003 001200 data division.
00004 001300 working-storage section.
00005 001400 01 n pic s9(18) comp.
00006 001500 01 bad-value pic s999 comp value -1.
00007 001600 procedure division.
00008 001700 p1.
00009 001800*********************************************
00010 001900* factorial(-1) with no size error phrase. *
00011 002000*********************************************
00012 002100 display "should give no size error phrase"
00013 002200 compute n = function factorial(bad-value)
00014 002300 display "test case did not abort".
When run, the following is output:
should give no size error phrase
No SIZE ERROR phrase (COBERR 747)
Program file: $OLDPASS.PUBS.COBOL74
Trap type = 80000000 (00,00), at pc = 0000035D.000053C7
assertion trap
DEBUG/XL A.47.01
HPDEBUG Intrinsic at: 118.00351270 cob_trap.print_message+$5c4
$$$$ Trap occurred at: trap_fact+$6c, at Stmt #13
PC=35d.000053c4 trap_fact+$6c
* 0) SP=4033a0c8 RP=35d.00000000
(end of NM stack)
============================================================
[REV END]
[REV
BEG]
In the next example, for a function that uses a procedure in the Compiler
Library, an invalid parameter is detected by the Compiler Library
routine. The Compiler Library routine generates the math library error
message and the COBOL trap handler provides the statement number of the
source line that produced the trap.
00001 000100$control post85
00001 001000 identification division.
00002 001100 program-id. trap-sqrt.
00003 001200 data division.
00004 001300 working-storage section.
00005 001400 01 n pic s999v999 comp.
00006 001500 01 bad-value pic s999 comp value -1.
00007 001600 procedure division.
00008 001700 p1.
00009 001800*********************************************
00010 001900* sqrt(-1) with no size error phrase. *
00011 002000*********************************************
00012 002100 display "should give no size error phrase"
00013 002200 compute n rounded = function sqrt(bad-value)
00014 002300 display "test case did not abort".
When run, the following is output:
should give no size error phrase
No SIZE ERROR phrase (COBERR 747)
Program file: $OLDPASS.PUBS.COBOL74
Library trap type = 0000000B (11)
DEBUG/XL A.47.01
HPDEBUG Intrinsic at: 7a0.00016ec4 cob_trap_lib.print_message+$158
$$$$ Call occurred at: trap_sqrt+$68, at Stmt #13
PC=7a0.00016ec4 cob_trap_lib.print_message+$158
0) SP=4033a480 RP=7a0.00017014 COB_TRAP_LIB+$a8
1) SP=4033a428 RP=7a0.0000d504 COB_TRAP_LIB+$8
export stub: 109.00377d14 FTN_GETLIBTRAP+$6c
2) SP=4033a250 RP=109.00377c74 ?FTN_GETLIBTRAP+$8
export stub: 109.005b1ddc DINVALIDERR+$e4
3) SP=4033a1d8 RP=109.005b12bc _dsqrterr+$1c
4) SP=4033a158 RP=109.005b0c98 label2+$10
5) SP=4033a128 RP=109.005b0b84 ?FTN_DSQRT$+$8
export stub: 308.00005398 trap_sqrt+$68
* 6) SP=4033a0d0 RP=308.00000000
(end of NM stack)
============================================================
*** MATH LIBRARY ERROR 28: DSQRT(X): X < 0.0 OR X = NaN
[REV END]
[REV
BEG]
The next example shows an IEEE floating point error.
00001 001000$control post85
00001 001100 identification division.
00002 001200 program-id. trap-float-zero.
00003 001300 data division.
00004 001400 working-storage section.
00005 001500 01 n pic s999v999 comp.
00006 001600 01 bad-value pic s999 comp value zero.
00007 001700 procedure division.
00008 001800 p1.
00009 001900*********************************************
00010 002000* divide by zero with no size error phrase. *
00011 002100*********************************************
00012 002200 display "should give no size error phrase".
00013 002300 compute n rounded = 1 / function sin(bad-value)
00014 002400 display "test case did not abort".
When run, the following is output:
should give no size error phrase
No SIZE ERROR phrase (COBERR 747)
Program file: $OLDPASS.PUBS.COBOL74
Trap type = 00020000 (14,27), at pc = 0000035D.000053B7
ieee divide by 0 DIV
operand 1 = 3FF0000000000000
operand 2 = 0000000000000000
result = 47C3500000000000
DEBUG/XL A.47.01
HPDEBUG Intrinsic at: 118.00351270 cob_trap.print_message+$5c4
$$$$ Trap occurred at: trap_float_zero+$7c, at Stmt #13
PC=35d.000053b4 trap_float_zero+$7c
* 0) SP=4033a0d0 RP=35d.00000000
(end of NM stack)
============================================================
[REV END]
Trace Traps.
Trace traps are "global" breakpoints that are easy to set up and use,
providing a very useful tool for quickly isolating program problems
within Debug.
Trace traps are global in that they apply to all paragraphs and sections,
entry and/or exit points. The COBOL compiler generates them within an
object program, and once generated, they can be armed and used with Debug
on MPE XL.
To use trace traps with your program:
1. Compile your program with the control option SYMDEBUG. This causes
the compiler to generate trace trap breakpoints in your code.[REV
BEG] (Note that these breakpoints are not available if the
SYMDEBUG=XDB option is specified.)[REV END]
2. Arm the trace trap breakpoints with the Debug command TRAP (see
the syntax that follows).
Syntax.
Type the underlined part in response to the Debug prompt.
$nmdebug> TRAP [trap_name] [option]
Parameters.
trap_name One of the following:
BEGIN_PROCEDURE Stops the program at the entry
to each procedure. For a
COBOL program, "procedure"
means program.
END_PROCEDURE Stops the program at the exit
from each procedure. For a
COBOL program, "procedure"
means program.
LABELS Stops the program at every
section and paragraph.
EXIT_PROGRAM Stops the program at program
exit point. For a COBOL
program, this is usually the
STOP RUN statement.
ENTER_PROGRAM Stops the program at the
program entry point.
TRACE_ALL All of the above.
option One of the following:
LIST List the current setting of
trap_name. This is the
default.
ARM Arm the trap specified by
trap_name.
DISARM Disarm the trap specified by
trap_name.
Example.
If you type the underlined command in response to the Debug prompt, the
program stops at every section and paragraph.
$nmdebug> TRAP LABELS ARM
NOTE [REV BEG]
The SYMDEBUG=TOOLSET control option significantly increases the
amount of object code and execution time[REV END] for your program,
so use it only while you are debugging your program. When the
program works, recompile it without SYMDEBUG.