New download vs "original" on disks

[Frank Cox]

When I originally purchased my PB/DOS 3.5 (upgrade from the previous version), the PDF manuals didn't yet exist. Therefore, I took Tom's advice in an earlier thread and requested a set of PDF manuals from [email protected]

I just received the link and downloaded the manuals. Thank you very much, guys! What a great way to keep a long-time customer happy.

Being the snoopy type, and since I got the whole DOS setup and installation program along with the download, I installed this new download to see if there were any differences.

Lo and behold:

PB/DOS on "old" 3.5" disk:
PB EXE 350,427 12-07-97 3:50a
PBC EXE 256,709 12-07-97 3:50a
PBD EXE 234,239 12-07-97 3:50a
PBINST EXE 78,832 12-07-97 3:50a
PBLIB EXE 46,870 12-07-97 3:50a

"New" download today:
PB EXE 350,827 12-19-97 3:50a
PBC EXE 257,109 12-19-97 3:50a
PBD EXE 234,399 12-19-97 3:50a
PBINST EXE 78,832 12-19-97 3:50a
PBLIB EXE 46,918 12-19-97 3:50a

I note that the filedates are different by 12 days, and the "new" files are slightly larger than the "originals".

The obvious question at this point is what changed, if anything?

------------------
PB/DOS 3.5 on Fedora Linux

[Michael Mattias]

UNOFFICIAL

The 'day' part of the file date represents different "builds" of the same release.

For example my files are dated 12-15-97.

Never seen any official document explaining what changed between builds, although I do have one document I got packaged with my release (which was an update sent to me after initial installation); the file name is "Fixes." and includes 'build' info through build 036. Not that it really tells me much: it's written like a document which was for internal use only and shouldn't have been sent to me.

[Frank Cox]

This "newer" PB compiles to a very slightly larger executable. A file that compiled to exactly 105000 bytes with the "old" compiler comes in at 105080 bytes with the newer version.

There is something different going on under the hood there.

------------------
PB/DOS 3.5 on Fedora Linux

[Fraser Portwood]

That's intreasting have just checked my PB35

PB.EXE 350,731
PBC.EXE 257,013
PBD.EXE 234,399
PBINST.EXE 78,832
PBLIB.EXE 46,918

All show a compile date 12-17-97

so it would appear that I have a build in between your original
and the one you just downloaded.

So it would appear that there where some fixes in the earlier
but not so important as to tell anyone.

Fraser


------------------

[Frank Cox]

I have discovered something rather interesting here.

As the "12-19-97" version of the compiler is apparently newer (and perhaps better) than the original one that I had (it even says "copyright 1999" when you first load it) I decided to start using the newer version to compile my projects.

That was apparently a mistake, in at least one case.

My clients get "Error 70 at pgm-ctr: 0" at seemingly random times when saving new records to their database when their program is compiled with the "12-19-97" version. When I compile the exact same thing using the "12-07-97" compiler, that error goes away.

I have switched back to the 12-07-97 version of the PB/DOS compiler to compile their program and they are a whole lot happier.

------------------
PB/DOS 3.5 on Fedora Linux

[Tom Hanlin]

It's so rarely a good idea to simply ignore an error. You might want
to track that down. The odds are pretty good that there's a bug in
your program that isn't as obvious in the older version of PB/DOS.

------------------
Tom Hanlin, PowerBASIC Staff
Opinions expressed may not be those of my employer or myself

[Frank Cox]

Tom, I agree completely. Unfortunately, I don't know where to start solving this problem, and I suspect that I don't have the information that I would require to nail the problem down.

Briefly, the problem is this: When compiled with the "new" version of the PB/DOS compiler, the program crashes at unpredictable intervals when saving a new record (adding a file to the end of the current database). When it crashes, it gives the error "Error 70 at pgm-ctr: 0". Trying to restart the program after this crash gives further crashes with more "Error 70 at pgm-ctr: 0" messages and the program doesn't appear to start running at all before crashing. It certainly doesn't access the disk at that point because the first thing the program doesn when it's loaded is present a menu on the screen and we don't get that far.

After an unpredictable number of retries loading the program (anywhere from 1 to a dozen or so) the "Error 70 at pgm ctr: 0" error goes away and the program starts and runs again just like nothing had happened.

There are two noteworthy features to this error.

First, when the program is compiled with the "old" compiler version, this error never (ever) occurs.

Second, the location "pgm-ctr: 0" appears to indicate that the program is crashing before it starts. It looks like White Queen territory, frankly. During the initial record save when it crashes, the program shouldn't be at location "0" even though that's what is reported. And during the subsequent fail-to-start sessions, I've not even looked at the disk yet before the program apparently crashes.

Note that this program is not running in anything resembling a "standard" DOS environment. It is a multi-user database running under DOSEMU with FreeDOS on a Fedora Core 3 application server that currently has six thin clients on the local network and two remote VNC sessions, one running on Windows XP and the other on Fedora Core 3.

My wild guess, based on no real information at all, is that some change that was made in the PB/DOS compiler between the "old" and "new" versions doesn't interact properly with FreeDOS SHARE under DOSEMU. However, having no changelogs or other information about what was actually changed between PB compiler versions, I am not in a position to guess at the real cause of the problem. It may be an obscure bug in my database program, it may be a bug in DOSEMU or FreeDOS SHARE that the new compiler version brings out, or it may be a bug or incompatibility in one of the fixes that was made to the "new" PB/DOS compiler.

But that brings me full circle. The program appears to work perfectly when it's compiled using the "old" compiler and it never crashes with that error. And I don't really have a starting point to work from to try to find cause the error that I am aware of.

Suggestions welcome, of course.

------------------
PB/DOS 3.5 on Fedora Linux