Re: Another DOS oddity...

From: Greg King <greg.king5_at_verizon.net>
Date: Sun, 26 May 2013 14:49:42 -0400
Message-id: <A910DD0BDB484E9D9E6C48E78659F477@Kaiser>
From: "Julian Perry"; on Saturday, May 25, 2013; at 8:02 PM -0400
>
> As I had recalled, "Inside Commodore DOS" by Immers & Neufield
> confirms the bug. (I remember seeing it myself.)  Apparently, there are 2
> bugs.
> One: the "referred blocks" occasionally will give you track 18 (!).
> Two: if a B-A is given to an already existing block, the error channel
> will return the next free block, but allocate all blocks on the requested
> track (which is what you're seeing)!
>
> The fix they suggest is quite drastic. It involves M-R reading the
> BAM in DOS RAM into a string on the C64 -- doing the block-allocate --
> and, if the B-A returns "No Block" -- M-W writing the allocated block
> BACK to RAM, overwriting the damage done.

That book is wrong in many places.

For example, U1/U2 didn't replace the B-R/B-W commands.  They are in 
different categories:  U1/U2 work on disks, while B-R/B-W work
on "files" (nameless, random-access files).  U1/U2 move sectors, while 
B-R/B-W move records.  A sector that has a record in it has a
record-length byte at its beginning.  Your program shouldn't touch that 
byte; the DOS controls it.  That's why the "#" buffer
pointer starts at offset 1 instead of offset 0.

Olaf is correct.  The DOS pre-allocates each full track, as a precaution 
against blackouts (if the drive loses power before the
modified BAM can be written back to the disk, then any file data that was 
written is protected).  The disk copy is updated when the
DOS moves to a different track -- and, when a file is closed.

A fix isn't needed because the block commands aren't broken.  They do 
exactly what they are supposed to do.  The bug is in André's
program.  It doesn't open a "#" file -- and, more importantly, it doesn't 
close that file!

A program must open "#" _before_ any B-A/B-F commands are sent; and, close 
it _after_ all B-A/B-F commands are sent.  It doesn't
matter that no data is sent through that buffer.  B-A and B-F are block 
commands; and, block commands work with "files"; so, you
must handle them that way.


       Message was sent through the cbm-hackers mailing list
Received on 2013-05-26 19:01:56

Archive generated by hypermail 2.2.0.