Google
 

Trailing-Edge - PDP-10 Archives - BB-5255D-BM - 4-documentation/checkd.tco
There are 15 other files named checkd.tco in the archive. Click here to see a list.
TOPS20 Change Order Number 4.2079

Written by: R.ACE		 6-Nov-78 07:37:51

Edit checked: YES			   Document: YES
  TCO Tested: YES		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD

Related TCO's:
Related SPR's: 


Problem: NEED COMMANDS FOR TAPE INITIALIZATION IN CHECKD

Diagnosis: 

Solution: ADD "INITIALIZE" AND "SHOW" COMMANDS TO CHECKD
[REFER TO LABELED TAPE SPEC FOR DESCRIPTION OF THESE COMMANDS]
TOPS20 Change Order Number 4.2114

Written by: DBELL		 8-Dec-78 11:08:32

Edit checked: NO			   Document: NO
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD
   Routines Affected:	NXTPR

Related TCO's:
Related SPR's: 


Problem: 
CHECKD incorrectly says file pages are in the BAT blocks, and misses
some pages which are.

Diagnosis: 
At NXTPR, it gets the address of a bad place from a BAT pair, but
only uses 18 bits.  On an RPO6, a disk address can be larger than this.

Solution: 
If this is a new style BAT pair, use the full 27 bit address, instead
of the 18 bit address.
TOPS20 Change Order Number 4.2115

Written by: DBELL		 8-Dec-78 14:38:54

Edit checked: NO			   Document: NO
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD

Related TCO's:
Related SPR's: 	 12561


Problem: 
BAT blocks reported wrongly by CHECKD.

Diagnosis: 
In release 3A, the location HFLAGS was inserted in front of the
location LPPCYL.  But that was in the middle of a table which was
BLT'd from other tables.  Thus LPPCYL got the wrong value, and
the BAT block computations messed up.

Solution: 
Move the storage location HFLAGS to a better spot.
TOPS20 Change Order Number 4.2119

Written by: DBELL		12-Dec-78 10:37:58

Edit checked: NO			   Document: NO
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD

Related TCO's:
Related SPR's: 	 12308


Problem: 
If there are more than 12K lost pages, CHECKD will release pages
which it should not.

Diagnosis: 
Pages to be released are stored in a buffer in memory, then written
to a file.  But the buffer's maximum size is 12K, thus when CHECKD
creates the lost pages file, it writes out some random junk following
the table.

Solution: 
Don't make a file that is larger than the buffer size.  At FNDLS9-7,
make sure LOSTOT isn't larger than MAXLOS.  Also fix the check at
FNDLS3+20 or so (it was an off-by-one bug).
TOPS20 Change Order Number 4.2232

Written by: DBELL		17-Apr-79 16:17:05

Edit checked: NO			   Document: NO
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: YES

Program: CHECKD

Related TCO's:
Related SPR's: 	 12851


Problem: 
CHECKD will not mark all pages bad that it should.  Exact same problem as
the monitor in TCO 4.2231.

Diagnosis: 
See 4.2231.

Solution: 
Round BAT block addresses down to a page boundary.  This is at location
NXTPR+6.
TOPS20 Change Order Number 4.2379

Written by: GRANT		 8-Aug-79 11:44:57

Edit checked: NO			   Document: NO
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD

Related TCO's:
Related SPR's: 	 13360


Problem: BADLY FORMATTED OUTPUT TO TERMINAL

Diagnosis: IN SOME PLACES A LINE FEED IS USED INSTEAD OF CRLF.

Solution: REMOVE THE USE OF LF ONLY AND ALWAYS USE CRLF.
TOPS20 Change Order Number 4.2496

Written by: GRADY		27-Sep-79 15:50:38

Edit checked: NO			   Document: YES
  TCO Tested: NO		Maintenance Release: NO
Hardware-related: NO

Program: CHECKD

Related TCO's:
Related SPR's: 	 13219


Problem: 

Checkd uses a file called (structure)-LOST-PAGES.BIN to save its' info about
lost pages, and later to release them during the "RELEASE" command.  Unfortun-
ately, this means that it is possible to really mess things up by doing lots
of "CHECK" commands and then lots of "RELEASE" commands.  You can also get a
similar mess with a few other contortions of reality caused by the fact that
Checkd can leave older generations of this lost pages files.


Diagnosis: 

Mostly, people seem to want to go out of their way to cause trouble, but its'
nice to make things as fool-resistent as possible.


Solution: 

When we do a "RELEASE" command, and we're all finished, instead of just deleting
the file we're using, delete ALL generations of it.  This involves a small 
change to the RLSPG2 routine.