Hello Bob!

Oct 14 16:25 03, Bob Jones wrote to Bo Simonsen:

 BJ>  ...
 BS>> I bet if you look in to the storage procedure in 
 BS>> vm_run.c, you might be thinking, "is this right?"

 BJ> Ok.  I'll need to add that file to what I should look at over the 
 BJ> weekend.

Take a coding weekend :-)

 BJ>> Unforuantely, due to the amount of time I have for this, I have the 
 BJ>> needed patience....  I'm barely getting around to 
 BJ>> repling to the echo 
 BJ>> mail at this point in time.....

 BS>> Okay. That's kinda sad..

 BJ> With the start of the new school year, I have been busier (sp?) than 
 BJ> I expected.  

So you do actually teach?

 BS>>>> But QWK upload doesn't work well :( 

 BJ>>> Try local console (bin/max -k) mode.  If I remember correctly, this 
 BJ>>> gets around the problem with file transfers and will allow you to 
 BJ>>> test out .QWK format without the file transfer code working.  

 BS>>> The file upload doesn't work at all1

 BJ>> Ok.  I haven't tested this idea.....  

 BS>> Okay, gotta rewrite.. It works _sometimes_ if you're lucky.

 BJ> Save a copy of the upload that works.  And then a copy of a small 
 BJ> upload that doesn't work.  Then use OD to dump out the contence of 
 BJ> each file and verify that both files are what they should be (CRC 
 BJ> wise, etc.).  Once both files are verified as good, then we know that 
 BJ> the problem is with either the user's program used to upload the 
 BJ> file, the link between the user and the BBS (the TCP/IP path if not 
 BJ> using the local console, direct file read if using the local console 
 BJ> methoid I've mentioned), or with the Maximus BBS code.  I suspect a 
 BJ> number of possible sources of problems.  CR/LF is one.  Non-clean 8 
 BJ> bit data path is another.  Character interpretation - changes by the 
 BJ> user program is another.  

It should parse double CR, LF and NULL. Also known as Byte stuffing if it's
what I'm thinking of.

 BJ> It might be usefull to put a debug log in Maximus to log what the 
 BJ> character stream that Max received was, and then compare that with 
 BJ> what was sent.  Then we can determine if things like ASCII NUL or 
 BJ> ASCII DEL are causing problems, or a CR/LF issue is there, or if an 7 
 BJ> vs 8 bit TCP/IP path is part of the issue.

Bingo.

 BJ> This is just the start of my thoughts on debugging this issue.

And I follow them.

Regards,
Bo

--- Msged/LNX 6.1.2 (Linux/2.4.22-1-586tsc (i586))