Date:  26-Aug-94 20:33 PDT
From:  Michael Devore [71540,62]
Subj:  Next version available

-------- Forwarded Message -------- 

Subject: Next version available
Date:    13-Aug-94 at 01:52   
From:    Michael Devore, 71540,62

TO: Gregory Bailey,76615,1350
   
CC: Jonathan Bigwood,100023,3553
    Robert Butler,76660,1514
    Mike Gould,71756,3176
    David Harms,72341,446
    Jeff Kurzner,70023,1751
    Neal Lochrie,72607,1331
    RANDY ROBINSON,70404,1617
   
The next version of CauseWay is available for those who need it.  Basically it
corrects the problem with applications that have >64K of global data leading to
startup GPF.  An correction in the extender that could cause a GPF at startup
with really huge programs (when running into several megabytes) was made. The
only other change is that a GPF in the report routines has been corrected.

This version passes my final test with one the huge Clarion application here, at
least as far as getting to the Btrieve missing error message since I don't have
Btrieve to load.  Any further corrections will need sample object modules to go
along with the error report so the problem application can be linked and the
error duplicated.

Let me know if you want or need the latest version.

Michael
    


  


Distribution:

To: Jeff Slarve > [73501,1323]

 

Date:  26-Aug-94 20:33 PDT
From:  Michael Devore [71540,62]
Subj:  New beta's up

-------- Forwarded Message -------- 

Subject: New beta's up
Date:    07-Aug-94 at 04:02   
From:    Michael Devore [EXEC], 72241,2106

TO: Gregory Bailey,76615,1350
   
CC: Robert Butler,76660,1514
    Jonathan Bigwood,100023,3553
    David Harms,72341,446
    Jeff Kurzner,70023,1751
    Neal Lochrie,72607,1331
    Mike Gould,71756,3176
    RANDY ROBINSON,70404,1617
   
I've finally had the necessary large block of free time to track down
and eliminate the Clarion CauseWay problem which was causing jumbled
screens and probably other more subtle error situations.

A couple of other bugfixes along the way were implemented.

New version of beta is going out to those reporting problems or other
recent feedback.  If you don't get a copy of the new beta with this
e-mail, you'll have to request it of me to receive it.

There still seems to be a GPF problem that can crop up if you have a huge
program with global data >64K such that several global data segments
are created.  Not sure if the problem always exists under these
conditions or just some of the time.  Anyway, I'll be working on fixing
that next as I still have a sample program which can cause problems.
The original test program which had screen problems now works 100%,
however.

If you continue to have problems with the latest beta, I'll need to
know.  Incidentally, if you wish to demonstrate the power of a DOS
extender to yourself or others, and your program operates using
CauseWay, do the following.  First, make sure you have free extended
memory, at least 300K-400K.  Then eat up all of you conventional/DOS
memory (e.g. via WARPHOG) except for 200K.  No Clarion program normally
runs in 200K of conventional memory, overlaid or not, trivial or not.  Run your
DOS-extended Clarion application.  Should work fine -- CauseWay will use the
extended memory to run the application, swapping to disk for virtual memory if
necessary for large programs.  CauseWay's physical memory requirements are 300K
(500K minimum recommended for performance reasons), of which only 100-200K need
be conventional memory.

Any questions, drop me a note.

Michael
   


  


Distribution:

To: Jeff Slarve > [73501,1323]

 

Date:  26-Aug-94 20:33 PDT
From:  Michael Devore [71540,62]
Subj:  08/26/94 beta

-------- Forwarded Message -------- 

Subject: 08/26/94 beta
Date:    26-Aug-94 at 21:14   
From:    Michael Devore, 71540,62

TO: Gregory Bailey,76615,1350
   
CC: Robert Butler,76660,1514
    Mike Gould,71756,3176
    RANDY ROBINSON,70404,1617
    Jonathan Bigwood,100023,3553
    David Harms,72341,446
    Jeff Kurzner,70023,1751
    Neal Lochrie,72607,1331
   
Next Clarion CauseWay beta is going out to those providing recent beta feedback.

This version should supply video selectors for DPMI, which means CauseWay linked
programs should now work under OS/2 and Windows -- provided that the direct
video addressing was the only problem being experienced.  A modification was
made to the main video routine in the runtime library and added to CWCLAR21.LIB

Another GPF fix was made to a runtime library routine and added to the
CWCLAR21.LIB replacement library.  Not sure exactly when the routine was called,
but it appeared to happen with one type of file I/O.

This version also contains the latest DOS extender and linker, which probably
doesn't affect any current tester.

Several LEMs break protected mode rules or are not CauseWay compliant.  I will
need a copy of those along with an example program that demonstrates the problem
in order to patch the BOJ's for proper operation.  I am supposed to be receiving
a Btrieve LEM and TSR soon for testing (I know what the problem is using
Btrieve, but have to figure out to fix it).  Communications LEM is reported to
fail but I will need an example program and BOJ to do any fixes there.

My own testing here has found two other Clarion third party LEMs which cause
problems, although they may not be fixed depending upon need.  The first is
SUPER.BOJ, don't know exactly what it does, would need an example to patch it.
The second is LEM_MGR which apparently compresses and automatically loads LEMs.
Since this is unnecessary with CauseWay which can hold all LEMs in memory at
once, this LEM will probably never be patched.

If you haven't received any recent betas or provided any feedback you are in
danger of being dropped from the beta team.  Please contact me at your earliest
convenience.

Michael
   


  


Distribution:

To: Jeff Slarve > [73501,1323]

 

Date:  26-Aug-94 22:46 PDT
From:  Michael Devore [71540,62]
Subj:  Oops again

While looking for a GPF that only occurred under DPMI, I uncovered a (probably
unrelated, but re-check your GPF's anyway) bug that's been in CauseWay forever.
Basically it's a serious memory leak.  I released a selector on a free memory
call, but not the associated memory.  This meant that each Clarion memory
allocation would eat memory that would never be returned.

Anyway, I'm sending out another CWCLAR21.LIB fast on the heels of the last one,
since this is pretty serious error even though it's been in there all during the
beta.  One person already reported a memory bug probably due to this guy.

Michael
  


Distribution:

To: Gregory Bailey > [76615,1350]

cc: RANDY ROBINSON > [70404,1617]       Mike Gould > [71756,3176]
    Jeff Slarve > [73501,1323]          Robert Butler > [76660,1514]

 

Date:  28-Aug-94 15:20 PDT
From:  Michael Devore [71540,62]
Subj:  Test patch

This linkable object module CWVM.OBJ is an attempt to get around a GPF that
occurs only under DPMI environments.  I can't tell if it works, because the
error reports I have are all CW.ERR or EXEcutable files so can't directly link
it in, but the change appears to address the problem.

Link CWVM.OBJ in the object module list OR add it via the configuration file
WARPLINK.DEF by the line:

CWVM=OBJADD
or
CWVM=OBJAPPEND

Since I can't check it very well, this patch may just move the GPF further down
the road, but let me know what happens.

Michael
   


Distribution:

To: Gregory Bailey > [76615,1350]

cc: Jeff Slarve > [73501,1323]          Robert Butler > [76660,1514]

 

Date:  28-Aug-94 18:22 PDT
From:  Bob Butler [76660,1514]
Subj:  WL32

 >> Can you please enlighten me to what kind of linker you were using
 >> before, and what kind of link script you were using?

I was using WarpLink 2.6.... there are a few tricks to WarpLink that are
important to understand before using WL32.  If you are not using Warplink 2.6,
I would recommend you get everything working with WarpLink first.  You might
want to download WLINK.ZIP from the Clarion CIS libraries (LIB 4 I think) to
learn about all I know.  There are 3 key things: 1) get your APPNAME.LNK file
working right, 2) Get WARPLINK.DEF files setup to match the latest version of
the BETA, and 3) Use the command line: WL32 /lc:WARPLINK.DEF @APPNAME.LNK.
I'll send you via CIS my LNK and DEF files.... maybe they'll give you a few
hints, even though they're for a different application.  Good Luck!

All the Best... Bob Butler, DATA.TXT Corporation

 

Date:  31-Aug-94 20:13 PDT
From:  Michael Devore [71540,62]
Subj:  Clarion beta 08/31/94

Here's the latest on the CauseWay for Clarion beta test.

1.  Error report with example on an environment manipulation bug in DOS1.BOJ
replacement.  Tested, duplicated, fixed, verified.

2.  CauseWay now supports the Btrieve LEM, or at least the example sent to me
for testing works.  I don't know if there are different LEM versions out there.
This was not a LEM patch, but a runtime patch to CauseWay to trap the Btrieve
interrupt and copy buffers down to and up from low memory so the real mode
Btrieve TSR would work correctly.

3.  As a side effect to the above, I changed how CauseWay handles LEM binary
calls.  The rest of this explanation is fairly technical, don't worry if you
don't understand a bit of it.  Previously a selector was aliased to the code
segment so that you could write to the code segment, as LEM's do.  Then the
alias selector was released so that they would not keep leaking and eventually
exhaust all available selectors as more and more of them were aliased to a code
segment.  The only problem with that approach is that some LEM's (like Btrieve)
keep a pointer to the aliased selector, and GPF when accessing it later on after
the selector has been freed.

I've changed the logic so that aliased selectors do NOT get released but are
placed in a table.  CauseWay checks this table before allocating another
selector for aliasing, supplying the same selector if it's already aliased to a
particular code segment.  This allows aliasing to occur normally without leaks,
but without later pointer problems to a freed selector.

I anticipate that this will clear up problems with other LEM's, as well.  Or
that's the hope, anyway.

4.  John changed the DOS extender code to gracefully handle OS/2's DPMI AUTO
setting.

Everyone later tonight or early tomorrow AM will receive a new ZIP containing
WL32.EXE, CWCLAR21.LIB, and the support files CWCL0.OBJ, CWCLAR.OBJ, and
CWDOS1.OBJ since most of them have been changed with the latest round of
revisions.  Those failing to provide any feedback within a week after
distribution will not receive further beta revisions.

Those new to the beta test will also receive the mishmash of documentation that
has collected over the past several beta revisions.  Check out README.OLD first
for an overview of what's going on and quickstart instructions, and work up from
there.

Following up on past reported news, as of August 26th Logix Development
Corporation has lost all rights to sell, upgrade, give away, or otherwise
distribute LPM-Link or variants, due to breach of contract.  Those of you
recommending linkers for purchase may wish to make note of the fact that the
Logix distribution path is now officially and legally closed.  That's probably
the most I should say on the matter at this time.

On a CauseWay related note, it appears that I may have to charge more for the
commercial release of the extender than I had originally planned and announced.
This is because the Clarion version has cost much more in development time and
effort than the other CauseWay versions, as well the fact that most of the
planned Clarion development and marketing budget vanished due to matters alluded
to above.  Therefore, it will be very difficult for me to recoup my costs at the
originally planned price of $145.00 (which is what the other CauseWay language
versions cost).  My current thinking is $245.00, with a substantial discount to
WarpLink owners during two month introductory rollout.  I'll entertain any
dissenting opinions any of you have on the matter, but do reserve the right to
be stubborn about it.

If you have any GPF errors with an application with this beta release, at a
minimum I will need the CW.ERR and MAP file.  There is a very good chance I will
also request a (small as possible) linkable OBJs example so that I can duplicate
the problem.

Michael



Distribution:

To: Gregory Bailey > [76615,1350]

cc: Jonathan Bigwood > [100023,3553]    Ragnar Hellspong > [100044,1214]
    John Wildsmith > [100116,2534]      Jeff Kurzner > [70023,1751]
    RANDY ROBINSON > [70404,1617]       Mike Gould > [71756,3176]
    David Harms > [72341,446]           Neal Lochrie > [72607,1331]
    Jeff Slarve > [73501,1323]          Ted Dircz > [74663,1563]
    Robert Butler > [76660,1514]



Date:  31-Aug-94 22:46 PDT
From:  Michael Devore [71540,62]
Subj:  Minor change

I backed up one DOS extender version since the latest just GPF'ed on me and the
earlier version didn't.  Just means that the OS/2 DPMI AUTO fix isn't in there
and that's about it.  Files are going out to everyone in the next few minutes.

Oh, it looks like Btrieve might be having problems reading information from its
data files, but I'm not real sure from the example.  Anyway, let me know if you
have problems with the Btrieve support.  As first cut, it's probably not
perfect.

Michael
   


Distribution:

To: Gregory Bailey > [76615,1350]

cc: Jonathan Bigwood > [100023,3553]    Ragnar Hellspong > [100044,1214]
    John Wildsmith > [100116,2534]      Jeff Kurzner > [70023,1751]
    RANDY ROBINSON > [70404,1617]       Mike Gould > [71756,3176]
    David Harms > [72341,446]           Neal Lochrie > [72607,1331]
    Jeff Slarve > [73501,1323]          Ted Dircz > [74663,1563]
    Robert Butler > [76660,1514]

 

Date:  07-Sep-94 21:57 PDT
From:  Michael Devore [71540,62]
Reply to: WL32

I don't know if Bob Butler's program runs okay, you'd have to ask him. The piece
he sent me for testing a long time ago has run for several months.

There will be a new beta with instructions later tonight.

Michael
  


Distribution:

To: Jeff Slarve > [73501,1323]

 

Date:  07-Sep-94 23:36 PDT
From:  Michael Devore [71540,62]
Subj:  CW beta

Next beta is ready and going out later tonight/AM.  Active testers on the list
are: Bailey, Butler, Gould, Hellspong, Robinson, and Slarve.  If you aren't on
this list, you need to contact me or you will be dropped from the beta program.

First, now is a good time to re-establish the ground rules for testing. Then
I'll discuss the beta changes.

If you have an exception/GPF error from the DOS extender, you must send me the
CW.ERR file and the application's MAP file current with the GPF'ing executable.
There are no exceptions to this rule; I have to have these files to do any
debugging whatsoever.  It does no good to report an exception error if you do
not send these files along.  It does no good to report that the program is
breaking in a particular spot, even if you think it is the same place as a
previous beta.  The only time an exception error is exactly the same from a
previous beta is if the CW.ERR file is exactly the same except for the copyright
version, the MAP file is identical except for time stamp, and you linked with
exactly the same version of object modules.

Referring me to a previous copy of a sample program sent along to duplicate the
error will not work.  Often I don't get the same error.  Frequently, my test
environment is completely different.  Later on, after the CW.ERR and MAP file
transfer, I may refer back to a previous sample program, but not for the initial
report.

There is a good chance I may request example files after looking at the
exception error or GPF, and a 100% chance I will request an example if the error
reported is not an exception or GPF.  For ease of transfer and bug tracking,
it's best to try and isolate a problem to a small example.  If that isn't
feasible or won't duplicate the problem then contact me and we'll see if we can
work out another way prior to send huge applications around to fill CompuServe's
coffers.  Example files consist of:  .OBJ and .BOJ files, any support data
files, and documentation on any required environment variable, directory, or
other system settings.

Currently I am receiving several hundred e-mails per month, as well as
performing normal business operations, getting the next newsletter ready, and
releasing an updated version of CauseWay for Clipper to a few hundred people.
Allow up to a week for a response to any beta report.  Of course, there may be a
response in a couple of hours, just don't count on it.

This beta incorporates two changes.  First, it uses the latest version of the
DOS extender, which corrects a problem under OS/2 DPMI AUTO setting.  Second,
I've made more changes to WL32.EXE related to the global data setting, since the
past changes could cause problems with file i/o.  It is my hope that these
latest changes will clear up problems several of you are having when files are
being read or written to.

The Btrieve situation, unfortunately, remains as last beta.  I've spent a lot of
time peering into the guts of the thing and it looks like all the low-level
interrupt stuff is working correctly in protected mode.  My test example,
however, still doesn't work quite right, not correctly reporting the presence of
data in a file.  My best guess right now is that there is a problem in the
higher level Clarion code, probably from a CauseWay nonfatal error, that messes
up a flag variable indicating pre-existing data in a file.  My alternate best
guess is that the internal Btrieve file positioning is still somehow
incompatible with CauseWay operation.  If you use Btrieve, try this beta version
and see if and where you have problems with it.

Michael

    


Distribution:

To: Gregory Bailey > [76615,1350]

cc: Jonathan Bigwood > [100023,3553]    Ragnar Hellspong > [100044,1214]
    John Wildsmith > [100116,2534]      Jeff Kurzner > [70023,1751]
    RANDY ROBINSON > [70404,1617]       Mike Gould > [71756,3176]
    David Harms > [72341,446]           Neal Lochrie > [72607,1331]
    Jeff Slarve > [73501,1323]          Ted Dircz > [74663,1563]
    Robert Butler > [76660,1514]

 

Date:  07-Sep-94 23:41 PDT
From:  Michael Devore [EXEC] [72241,2106]
Subj:  beta0908.zip


 % Message is Binary 

 DOW /PROTO:B

  3 Michael Devore [EXEC]/beta0908.zip
   07-Sep-94 23:41 PDT 72241,2106 Length 106976
    * Binary *

1 messages and 106976 characters ready for download
Enter a filename for your computer: CM090894.1

Downloading to CM090894.1


*** File Transfer Completed! ***

 

Date:  14-Sep-94 21:06 PDT
From:  Michael Devore [71540,62]
Subj:  CauseWay 09/14/94

Here's the latest on CauseWay for Clarion 2.1 news.

Straight Clarion code seems fairly stable, there have been no GPF reports on
pure Clarion code since last beta release.

LEMs are proving to be tricky.  Currently there are three modified LEMs for
protected mode: DOS1 (everyone should have this as CWDOS1.OBJ), COMM, and
ALERTED.  If anyone needs the final two let me know.  Some LEMs need no
modifications for protected mode, but apparently many others do.

Btrieve and Supermemo LEMs are prone to acting up under CauseWay, but not for
reasons I can detect at this time.

There has been a patch in the CWCLAR21.LIB library since last full beta, but
it's not generally required.  I changed the way the binary calling routine works
so that the GS segment register always points to the code segment.  This should
make things easier for LEM developers because they simply replace CS-relative
writes with GS-relative writes.  You probably need the updated library if you
request the CauseWay version of COMM or ALERTED.

I'll be getting together first rough draft on Clarion CauseWay for the
forthcoming manual in the next 2-3 days.  Let me know if you want to look it
over.  It's already slated to be delivered to the SuperMemo developer -- his
name escapes me as I write this missive away from my CompuServe address list,
Robert P. is all I can remember.  Anyway, he'll also get the full set of
CauseWay files for testing with his stuff.

At this stage, besides drafting the manual, I'm pretty much just slogging
through problem LEM's.  Let me know if you have any problems with anything.

The Slash i newsletter going out next week to all WarpLink and CauseWay owners
will mention that CauseWay for Clarion 2.1 is in pre-release, so if you want to
talk about it, feel free.  I'm going to make it available at $175.00
introductory price to non-beta regular customers on a case-by-case basis if they
request, assuming that they are not using a problem LEM.

Bottom line is that I know there is a problem with Btrieve and SuperMemo LEM
using CauseWay, but no other operational problem reports.  If you have anything,
you need to let me know.

Michael



Distribution:

To: Gregory Bailey > [76615,1350]

cc: Ragnar Hellspong > [100044,1214]    John Wildsmith > [100116,2534]
    Jeff Kurzner > [70023,1751]         RANDY ROBINSON > [70404,1617]
    ROBERT PUPAZZONI > [70441,204]      Mike Gould > [71756,3176]
    Jeff Slarve > [73501,1323]          Ted Dircz > [74663,1563]
    Brian Staff > [75310,1147]          Robert Butler > [76660,1514]



Date:  22-Sep-94 07:15 PDT
From:  ROBERT PUPAZZONI [70441,204]
Subj:  SuperMemo / CW32 Testing, cont.


Attached file: CWBINFNC.OBJ


 >> The LPM one, would allow the choice of keymaps, but the [ctrl_enter]
 >> would not save a form with a memo, and I could not escape out...

This is probably a result of the ALERTED() problem.

Michael's new CWBINFNC.OBJ will fix that.  I've attached a copy this message.

Add the following line to the end of your app's .DEF file:

 cwbinfnc=OBJADD

Bob


Distribution:

To: Jeff Slarve > [73501,1323]

 

Date:  23-Sep-94 13:12 PDT
From:  ROBERT PUPAZZONI [70441,204]
Subj:  SuperMemo / CauseWay32 Update

Attached file: SMSPELL.BIN


SuperMemo/CauseWay Beta Testers:

(1) Attached is an updated SMSPELL.BIN for SuperMemo 2.0x which should
    allow the spellcheck function to work properly in programs linked with
    CauseWay/WarpLink 32.

    (BTW, it will also work with Processor, RTLink, LPM VM-Link, etc.)

    It seems to work ok, but I have not tested it extensively, so let me
    know if there are any problems.

(2) Michael Devore has also confirmed a bug in the CW32 libraries which
    causes the SuperMemo ALERTED.BIN LEM to malfunction.  Please ask
    Michael to send you the file CWBINFNC.OBJ to correct this problem.

Thanks.

  Robert Pupazzoni
  Bobcat Systems


Distribution:

To: Ragnar Hellspong > [100044,1214]    Jeff Slarve > [73501,1323]
    Ted Dircz > [74663,1563]




Distribution:

To: JEFF > [73501,1323]



Date:  04-Oct-94 23:41 PDT
From:  Michael Devore [71540,62]
Subj:  CauseWay 10/05/94

Latest modification report on CauseWay for Clarion 2.1....

Same runtime modification made to ROWS() routine as was done to ROW() routine.

Added checks for another internal Clarion move routine to catch out of bounds
moves, as well as converted it to 32-bit instructions.  Speed increases with the
low-level routines normally would be greatly improved with the 32-bit changes
except I usually have to add several boundary checks because Clarion is terribly
sloppy about memory moves and copies, often shooting past an allocated buffer by
one or more bytes.

Added a couple more routine changes to better support global data >64K.  These
new changes catch problems when information used to call a LEM routine spans a
segment, automatically adjusting for the memory read into the next segment.
There may be other routines that must be modified in the future to allow for
information segment-spanning, depending on where a particular piece of
information falls in large programs.

The above three modifications allow operation of all example programs sent to
me, except for SuperModel error reports.  Apparently SuperModel LEMs are not
fully protected mode compatible.  To patch it/them, I'll need example programs
(preferably as small as possible) that demonstrate a GPF problem using one of
those routines.  Examples should include .OBJ and .BOJ files as well as the link
file.

Everyone will be sent the CWCLAR21.LIB library and the latest CWCLAR.OBJ to make
sure we're all in sync.  A couple of error reports were not duplicatable by me
and old versions may be the reason.

Michael
   


Distribution:

To: Gregory Bailey > [76615,1350]

cc: Jonathan Bigwood > [100023,3553]    Warren Marshall > [100033,3544]
    Ragnar Hellspong > [100044,1214]    John Wildsmith > [100116,2534]
    Jeff Kurzner > [70023,1751]         RANDY ROBINSON > [70404,1617]
    Bud Gallagher > [70413,3101]        ROBERT PUPAZZONI > [70441,204]
    Robert L. Sorrells > [71202,1525]   ONTOP > [71722,534]
    Mike Gould > [71756,3176]           Helmut J. Schwarzin > [73140,2732]
    Jeff Slarve > [73501,1323]          Lee G. White  [LSi] > [74067,3441]
    Ted Dircz > [74663,1563]            Brian Staff > [75310,1147]
    Robert Butler > [76660,1514]

 

