Re: FW: Re: 5250 Data Stream Operation Codes

On Tue, Dec 14, 1999 at 08:45:47AM -0600, wsuetholz@centonline.com wrote:
> 
> On 10-Dec-99 Jason M. Felice wrote:
> > On Sat, Dec 11, 1999 at 11:00:26AM +1300, Carey Evans wrote:
> >> > Out of curiosity, why would we do that?
> >> 
> >> Because if tn5250 crashes, and you reconnect and resume the session,
> >> the local buffers won't be there any more.  And if you signoff a
> >> session when it's got some screens saved, you get a memory leak since
> >> the saved screen is never used.
> 
> The memory leak is there, and there really isn't a good way to determine when
> a saved screen buffer could be freed.
> 
> > 
> > Waitaminute... what happens if you reconnect the session with a different
> > emulator?  That could be disasterous.
> > 
> 
> Yes,  And since the restore can possibly use the restore opcode, you can't
> just reverse engineer the screen and send the WTD type commands and expect
> another emulator to handle it properly.  However, that was the solution that
> I did.  Basically I determined the WTD orders necessary to repaint the screen
> and get my fields back to the same state they were in.  Modified flags and
> all.  Then I treated a restore opcode like a put/get opcode which left the
> session in the invited state.

I think that is the best way to go - it's also reusable later down the line
when I implement a 5250 'server' which uses the same API :)  That, and it
shouldn't be too difficult.

I'm still going to keep the 'multiple format table/display buffer' structure
of lib5250 for later when I implement a 5250 server as well - then I'll just
have to add fields to the display buffer and format tables which keep track
of the client-supplied save data and we're all good.

> 
> As far as other emulators go...  Shouldn't everyone be using the better tn5250
> emulator for linux?  ;-) 

:)

On another note, I think Sean Porterfield submitted the bug with transmitting
negative numbers in Number Only fields.  I've been looking at this, and I can
duplicate it (easily, it appears to happen in every case), but a very close
inspection of the logs seems to indicate that I'm doing nothing wrong.  I've
also closely inspected the 5250 Functions Reference manual, and it doesn't
indicate that the data from this field should be transmitted in a special
manner.  I'm going to try snooping on a session with a different emulator
if I get a chance at lunch or before I leave work today; regardless, a log
from 'portsnoop' (included in the tarball) with an emulator that you know
works would be terribly useful.

This is a big deal from my perspective, and I'd like this to work properly
in the next development release (as well as the save/restore issue above).

Also, I've been having a particular problem with the version of the emulator
installed at work (but can't duplicate it at home)...  I seem to recall
someone else having a similar problem but I couldn't duplicate it at the time:
If I go into strpdm and take an option 3, the format table and the display
don't seem to be properly aligned, meaning that I have to enter data one
character to the right of where it should be entered.  Is anyone still having
this problem?  If so, can you post some information on your system
(what distribution, version of libc, version of curses, output of 
"ldd `which tn5250`", version of autoconf and automake if you got sources
from CVS, and the version of the emulator).

Thanks,
-Jay 'Eraserhead' Felice

+---
| This is the LINUX5250 Mailing List!
| To submit a new message, send your mail to LINUX5250@midrange.com.
| To subscribe to this list send email to LINUX5250-SUB@midrange.com.
| To unsubscribe from this list send email to LINUX5250-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---