Treatment of Tape and System Errors

R Brandwood

11 October 1975

Faults can be divided into 3 main groups:

1. Tape faults: these are normally reported as TAPERROR or LONG RECORD.
2. Error messages: these are reported by the DISPLAYER programs (FRFO etc ) as ,*UNC,*DLM etc.
3. Others: these are those not falling into categories 1 or 2 - example, jobs returning to DEBUG during execution.

1. TAPE FAILS

These are, or should be, reported on the teletype as TAPERROR, preceded by the tape order and tape status word printed in octal.

Action to take:

1. Type TRYAGAIN/20. This will cause the block to be reread 20 times or until successfully read, whichever happens first.
2. If error persists, then terminate job. DO NOT type CONTINUE or GO. Use the SKIP/ command to position tape at trailing SYSLOG, return contral to MON;SYSLOG and type GO/. This will allow the trailing logging job to be completed and other jobs (if any) on the tape to be completed. The failing job should be recorded as such.
3. Use the 1906A or the PDP15 to copy this tape and rerun the failing job. If all attempts to nm the Job fail, then return the output to the user accompanied by one of the attached forms.

Long Record

This usally indicates a poor tape, or dirt on the tape drive.

Action to take:

1. Terminate as in previous para (b).
2. Have the tape and drive cleaned.
3. Rerun the job. If error persists then copy the tape and try again.
4. If the error still persists, then treat as an "OTHER" type of fault.

2. ERROR MESSAGES

These errors are preceded by '*' - example *DLM. These can be caused by a tape fault, or by the user, or by the graphics package writing the wrong orders to tape. A list of these errors can be found in the COMP80/FR80 User Manual, page 84.

Action to take:

1. Reposition the tape at the start of the graphics data using the BACK/ and SKIP/ command. BACK/ will reposition tape and last filemark. This command will have to be repeated if a filemark has already been dealt with by the DISPLAYER. SKIP/2 will then skip the 2 records used for LOADGO. (GO/ will restart the job.) If the error reappears, then the job should be terminated as in para (b) on page 1.
2. The output and details of the fault should be passed to the user, use a copy of the attached form to convey details.

Example of error and values preceding:

777777 - to start console teletype 
40001  - failing data word 
1      - record number 
88     - word position in record 
*DLM   - type of error 

3. OTHER FAULTS

These are faults not covered by 1 and 2. Rerun the job. If the error persists, the following should be done:

(a) Return control to DEBUG      CTRL-D
(b) Load program with symbols    filename $S
(c) Find the value of            FILNUM/
                                 FILNUM-1/
                                 CURBUF/
                                 MTPTR/
                                 or 14/

If response to these is not a numerical value then type a : to obtain one.

Using the value of CURBUF print out 400 words of the program in the following way:

$M+1/ insert value of CURBUF
$M+2/ insert value of CURBUF + 400 (these numbers are octal) 
720000$N 

The teletype will then print out 400 lines of program. This should take approximately 20 minutes. The printing can be stopped by depressing the space bar.

One copy of the output and teletype listing is to be passed to Eric Thomas as soon as possible.

The other copy of output is to be returned to the user accompanied by a copy of the attached form informing him that an error was detected in the job and is being investigated.

Example:

FRFO

CAM E16*OK
PULL3150*OK
X NUM2*OK
Y NUM2*OK
00

                          ** START JOB   **   16:49'47/.9"

*DEBUG
FRFO$S
FILNUM/2
FILNUM-1     25
CURPUF/  ..TMP4+3     :16304
MTPTR/?
14/      ..END    :16305
$M+1/   0      16304
$M+2/   37714  16704
720000$M
..TMP4+3            JMS  I  FDSFSS+4
END
..END,    JMS I FDSFSS+4
          DAC FDINIK+47
          AND I FDIZMP+6