Subj : Grunged message cap pointer help?
To   : Mark Lewis
From : Mike Luther
Date : Wed Sep 05 2001 04:01 pm

And notwithstanding my previous post Mark ..

ML> I took three messages in the NETMAIL directory and using a
ML> command line session in a window in native OS/2, simply
ML> renamed the messages from lower case to upper case as in:

ML>          ren *.msg *.MSG

ML> Poof!   No more run-away message counter!

ml> somehow i don't see it... it's never done this on my
ml> box over here... what OS are you running, again? yeah,
ml> OS/2 but what 'flavor and fixpaks?? disk drivers and
ml> such...

That too, still may be a part of the solution!

Here is why.  This box is a complete SCSI box on Warp 4.0 FixPack 15.  It has
the IBM authorized kernel patch of WR0508.ZIP on it.  Has the complete MPTN
8423, 8424 and 8425 fixes, as well as UN980 for TCPIP 4.1 which is still on it,
together with the PEER fix 8413, IIRC.  There may be a later PEER fix on it.

I have discovered that CERTAIN DOS applications, when running under a DOS-VDM
in OS/2 are actually leaving LOWER CASE letters in the file names! Not only
that, but under DOS-VDM's in OS/2, if the application was written in some
compilers and they compiler was actually writting the lower case name to the
disk, you've got it!  The file name is actually being accepted and handled in
lower case by OS/2, even though a conventional pure command line DIR *.*, for
example, in DOS, translates it to UPPER case as always!

Now, if you take the OS/2 command line session, and do that DIR *.*, you and
see both the UPPER and LOWER CASE file names there!  Of course, WIN-95 and OS/2
native code would and could and do that!  But those programs which were written
by DOS programmers a while back which did not earlier have any problem with
this because DOS gave you back, or mostly did, all UPPER case,would forgive you
of your sins come some other open or close attempt!

Early on we who used the MicroSoft PD7 BASCOM compiler had to watch for this in
certain DOS operating systems, including, it seems, OS/2 16 bit early stuff!
There was a caution in the handbook about being sure to use all UPPER CASE file
names in your work, because PD7 could actually create and open and close a file
like, "MyFiLe.TxT" and, if you did it, only, at that time, PD7 could see it and
handle it.  DOS itself, and most other programs,could not handle that at all.

Of course the advent of WIN-95 changed all this.

I think what is happening in the bug I've seen, is partly related to this,and
as well partly related to the other portion of the 'fix', I noted in the
previous message.

As long as there is ONE message, MSG #1, on the system in NetMail, which was in
INBOUND message to the system, and stays there, read or un-read, this error
never seems to show up on this box again!

ML> As far as I think I recall, FASTECHO is written in Borlund
ML> Turbo Pascal,

ml> flavor
ml> ====================
ml> DOS      Borland C++ - Copyright 1991 Borland Intl.
ml> OS/2     Borland C++ - Copyright 1994 Borland Intl.
ml> 16bit    Borland C++ - Copyright 1993 Borland Intl.
ml> NT       Borland C++ - Copyright 1993 Borland Intl.

OK .. still very much in vogue for non-GUI work with those who know the
language well, I think.

ml> all that i have on my system are currently dated 03
ml> APR 1997 except for the NT version which is 01 JAN
ml> 1997...

ml> again, i still think that it's due to corruption of
ml> the program code in the EXE... at least look in that
ml> direction and maybe do a binary compare out of
ml> curiousity... tell ya what, send me a copy of that exe
ml> file and i'll take a look at what i can see...

Does my explanation make any sense?

I know that I'm hitting all kinds of problems in FILE and MESSAGE area .CTL
files in MAX with this.  I have to be very careful.  The HUB was an old DOS
implementation with all UPPER CASE path names!  I've been a lower case guy in
describing them for years.

In mixing and integrating the two systems, I've discovered that by golly some
of the programs in the DOS camp are making irregular errors with the paths
which are in LOWER case which were created by the SILTP command for MAX.  If
you used SILT, the DOS version, they are UPPER CASE, even though your
description was a lower case one!  If you use SILTP, it even makes them LOWER
CASE and .. to some progams, the program cannot handle that!

I'm almost of the opinion after working carefully with the beta code on these
upgraded kernels for OS/2 from Austin here, that we are starting to see more
problems in this post original FP 15 kernel code.  I don't think anyone has
seen this yet and reported it.  I suspect that's because so few people in the
for-real business world of OS/2 are still using DOS.

The bad part of this is that I don't have any actual way of proving it up,since
I don't have the source for these programs that are hitting it. Therefore, it
never got reported or asked of Austin during the massive work that was recently
done on the Warp 4.5 kernel for MCP and ACP and FP 15 as well.  I'f I'd seen it
or know it, I'd have asked about it.  By now Austin has moved forward on other
things and I've not seen a later kernel than about August 8th or so.  That one
had no mention of anything like this in it at all....

    gloom.

Mike @ 1:117/100

---
* Origin: BV HUB CLL(409)696-3600 (1:117/100)