Problems handling chain offsets with OS/2's WriteAndX/ReadAndX

[samba.10.maazl at spamgourmet.com]

On 07.08.11 04.10, Günter Kukkukk - linux at kukkukk.com wrote:
> which OS/2 application are you using to tag mp3 files?
> I could install it here and with a short explanation from you how
> to use that program to get to the failing state, i could have a
> closer look.

It is an not yet released version of PM123. And it only happens if I 
edit an mp3 tag in a way that the ID3V2 tag is updated in place and the 
ID3V1 tag is created. Furthermore the caching strategy of IBM requester 
is relevant. If I try too often, the error seems to vanish until the 
requester is restarted. From the logs I can see that the requester 
changes the SMB command sequence to use easier commands if it got too 
many errors. Maybe essentially the reconnects because of the restarts of 
smbd to change the log level trigger this behavior change.

So I could provide the required executables and the steps to reproduce, 
but I am not that sure, that your setup will behave identical. It took 
me about one week to figure out a reproducable test case.Up to that I 
had only occasional error messages.


I reduced the test case by a special test executable, created from 
PM123's logfile. It contains all I/O calls to the file handle that 
causes the problem. At least on my machine this immediately triggers the 
problem at the last DosRead command as soon as it is run from a samba share.

home.arcor.de/maazl/temp/sambatest.zip


Marcel