Super-classed controls and cbWndExtra..

[Borje Hagsten]

Did some tests on standard and super-classed standard controls right
after creation with:

  MSGBOX STR$( GetClassLong(hCtrl, %GCL_CBWNDEXTRA) )

..with interesting results. BUTTON controls already has 8 bytes allocated,
COMBOBOX 2 bytes, etc. When Super classing them, one often want to add a
few bytes to cbWndExtra for own purposes, but since MS obviously already
use some of it and also say we should not set cbWndExtra to more than 40
extra bytes - then how safe is it to add anything extra at all?

I mean, what if MS starts using up those 40 bytes itself in later versions
of Windows? Maybe adding to cbWndExtra should be avoided with super-classed
controls? Or maybe I simply have been sitting here too long and my mind has
started seeing ghosts in every corner of Windows..


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

[Chris Boss]

Borje;

That is a good point !

While I doubt the standard controls will be modified significantly
in the future, the common controls obviously have gone through
significant changes because of many different versions.

Using properties may be a safer way to go.

I superclass the edit control which uses 10 bytes I think and I was
worried about that.

It does seem the standard controls don't use anywhere near the
maximum of 40 bytes though, so adding a few extra bytes shouldn't be
a problem. Also Windows 95 is 16 bit under the hood, so it is
limited by 40 extra window bytes, but Windows NT, ME, 2000 should be
32 bit under the hood and they may not be limited by the 40 byte
limitation. Have you tried Windows ME, 2000 to see if it will handle
more than 40 bytes ?

If it does, then all you would have to do is verify that Win95/98
leave you some room to work with and not worry about later OS.


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

[Borje Hagsten]

"Shouldn't" and Microsoft logic doesn't rhyme very well in my ears.
I think you are right Chris, probably nothing to worry about. If MS
starts using up lots of those bytes, it will be in new systems, new
controls, where limit is set higher. Just as long as one makes sure
to *add* own extra bytes to existing, it should be safe.

Still, the phrase "no one will ever need more than 640 KB" keeps
ringing in my ears.. :/

I have a fancy box with expensive Win ME here, no longer installed,
because I ran into so many hardware problems. MS seems to think
hardware older than a year doesn't exist and that all people are
rich enough to buy new each year. I'm a rare creature in their eyes,
some of the stuff I use daily is more than ten years old (like an
18-pin invoice printer from around 1848, etc.. So, can't test in ME..


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

[Steve Hutchesson]

Borje,

You are right, there are ghosts in every corner of windows and
Microsoft put them there. I am a "s"u"c"k" it and see man(empirical
testing) so if you need the extra bytes for each control instance,
write the code and get a few friends to test it on later versions
for you to see what it does.

You are probably safe to allocate the full 40 bytes and use the
higher addresses. I would not be surprised what Microsoft does
after seeing the changes from win3.## to win95 so I don't think
there is a particularly safe way to do anything as documented
functions get changed by Microsoft as it suits them, not you
or I.

Regards,

[email protected]

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


[This message has been edited by Steve Hutchesson (edited May 12, 2001).]

[Edwin Knoppert]

At the end, the controls are prob. not meant to superclass in the first place.
Superclassing is just made possible.
I never touch the extrabytes during superclassing.
You can not predict the results.
Maybe now but what in the future for that class?

SetPROP() is what i use.
Superclassed controls are easy to maintain because of the messages you get like WM_DESTROY.



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

[Semen Matusovski]

Why to worry ?
limitation of 40 bytes is for Windows 95 only (which almost died).
If OS doesn't accept large parameters it's honestly returns 0 in RegisterClass (so, possible to control).

If you need more than 40 bytes, you can:
1) either to mark an app by Windows 98+, NT 3.51+ (doubt that you lose more than 1% of customers)
2) either to use a way, recommended by Microsoft - 4 additional bytes only (pointer to allocated memory).

Actually second way is even better:
a) one GetWindowLong only.
b) clear that all standart controls use less than 40 bytes.

------------------
E-MAIL: [email protected]

[Chris Boss]

Borje;

I would assume that most controls will leave enough room in the
extra bytes to allocate at least one extra long.

To play it safe, you can use just one long to store a handle to a
UDT stored in Global memory (use Peek and Poke to move the memory
back and forth between a real PB UDT variable - of course you must
use the Global handle and unlock it first to use Peek and Poke).

This will allow you to allocate as large a block of memory you
may need.



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

[Borje Hagsten]

Why worry? Because soon new version of Windows again. Who knows what
MS starts doing there - maybe new limit will be 2 GB, but MS will use
it all for new .NET things in standard controls. Remember Bill's
prediction about 640 KB.. :/

No, I don't think so either. Just wanted to point out that not all bytes
are available when one super-class existing standard controls. I usually
allocate memory for a UDT and only need to store a 4-byte pointer. Should
be safe enough. However, did a stupid thing, superclassed a ComboBox and
tried to use Get/SetWindowLong at byte zero for my 4 byte pointer, with
GPF as result. Forgot that MS already use some of those extra bytes in
their controls.

BTW, did you know that dialogs, like DDT or resource, uses up 16 of those
extra bytes themselves? SDK style windows use what we tell them to, plus
gives programmer full control. Start to like SDK'style better and better..

BTW2: Win95 is far from dead. Millions of people have not upgraded, because
paying so much money just to get new start-up picture is not on their
priority list. As a matter of fact, lots of schools and even big companies
still even use Win 3.1(1) because A: it works for them, B: they have hundreds
of working old 486's and simply can't afford to upgrade them all. I know, I
still sell a lot of my 16-bit app' to both schools and companies and almost
not a day without a system manager that tells me how lucky he was to find
something that works in all his computers..

Recent statistics I saw in a newspaper showed about 80% use Win98/ME. Almost 10%
of all Windows systems were still Win95 (a lot of computers, that's for sure),
a few percent are still Win3.x and only a few percent actually is Win2K. Mostly
large companies use Win2K or NT4, but most ordinary users don't. So, depending
on type of program and market, it may be a waiste of time trying to figure out
things that don't work in a few % NT systems, but works in 90% of all PC computers..


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

[Borje Hagsten]

Chris, almost exactly what I do. In this case, something like:

'in declares
TYPE CBCOLORDATA
  oldProc   AS LONG
  autoColor AS LONG
  userColor AS LONG
END TYPE
 
'in WINMAIN or PBMAIN..
  LOCAL cbc AS CBCOLORDATA PTR
  cbc = HeapAlloc(GetProcessHeap(), %HEAP_ZERO_MEMORY, SIZEOF(@cbc))
  IF cbc THEN
     @cbc.oldProc   = SetWindowLong(hCombo, %GWL_WNDPROC, BYVAL CODEPTR(CBProc))
     SetClassLong hCombo, %GCL_CBWNDEXTRA, 40 'what is 40 bytes in ocean of memory..?
     SetWindowLong hCombo, 36, cbc            'Store the pointer for later use at pos 36
  ELSE
     EXIT FUNCTION                            'no point to continue..
  END IF
 
 
'in cbProc
  LOCAL cbc AS CBCOLORDATA PTR
  cbc = GetWindowLong(hWnd, 36)   'Get stored data
 
  SELECT CASE wMsg
     CASE %WM_DESTROY 'Un-subclass Combobox and free up allocated memory
        IF @cbc.oldProc THEN SetWindowLong hWnd, %GWL_WNDPROC, @cbc.oldProc
        IF cbc THEN CALL HeapFree(GetProcessHeap(), 0, BYVAL cbc)
        FUNCTION = 0 : EXIT FUNCTION

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

[Eric Pearson]

Borje --

> So, depending on type of program and market,

That's exactly right. For example, one of my programs is used by businesses that use 100.00% Windows NT. So it's a waste of time for me to do anything with 9x when it comes to that program. It all depends on your target audience.

> it may be a waiste of time trying to figure out
> things that don't work in a few % NT systems, but
> works in 90% of all PC computers.

Unfortunately that may turn out to be short-sighted. Microsoft has officially terminated the Windows 95/98/ME line. It is still available -- you can still buy 98 and ME -- but the replacement for those versions will be Windows XP, which is based on Windows NT.

It may be true that today a large percentage of the world's PCs use Windows 9x, but that number is virutally guaranteed to decline over the next few years.

-- Eric




------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>

[Borje Hagsten]

Things will be very different with XP, because most people will have
to upgrade or buy new computers in order to run it. Plus, more than 90%
of all Internet connections are still via slow modems, and will be for
many years to come, so most people in real world won't be able to use
XP and many of those .NET features at all.

Must remember that for majority of people, a computer is not the most
important thing to spend money on. I have a feeling MS marketers have
fallen into latest IT hype and think all has broadband and GHz machines,
or at least will get/have within reasonable time. Have a feeling that
putting so much trust and money on XP may be what finally pulls Microsoft
down from the throne.

Also have feeling they are not complete fools, so must be some weird
strategy behind all this. Maybe they want "ordinary" people to start
using Mac instead? After all, Bill has always had his greedy eyes on
Mac and owns a large portion of it by now, so who knows..


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

[Eric Pearson]

Borje --

Remember, just 15 years ago the standard system was a 640k XT (an "8086") running DOS, with a mono screen, a 20 meg hard drive, a 5-inch floppy drive, no modem and no mouse.

Ten years ago most people were using using 286 or 386 systems, with DOS or maybe Windows 3.0.

Five years ago Windows 95 was just a few months old, and a 200 Mhz Pentium Pro was something you invited your friends over to see.

Today we have 1 GHz machines with multi-gig hard drives, Windows ME and 2000, and the growth of high-speed internet connections is following the same curve that everything else has followed.

Five years from now? I don't know if .NET or XP or anything else will still be around, but I am certain that things will be different than they are today. So unless you are writing software that you intend to have a lifespan of only a few years, IMO it is a mistake to base your plans on the current state of the world.

So when I read...

> more than 90% of all Internet connections are
> still via slow modems, and will be for many years
> to come, so most people in real world won't be
> able to use XP and many of those .NET features at all.

...it sounds very much like somebody saying "640k should be enough for anybody". (Sorry, couldn't resist. )

-- Eric


------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>



[This message has been edited by Eric Pearson (edited May 13, 2001).]

[Steve Hutchesson]

I am incline to agree with Borje about the average hardware
around the world, outside of some European countries and the US
bandwidth is expensive and unreliable and slow. try running streaming
video on a 28k connection and you will understand what the problem
is.

High end hardware belongs to the corporate sector and many developers
but assuming that smaller businesses and the consumer market have
the same is a mistake, many machines are not connected to the internet
at all as the users have no use for it.

Price fatigue is another factor, small business and the consumer
market generally object to repeatedly paying very high prices for
software that often does not work as well as their older stuff.

I deal with programmers around the world and in many places, a 133
meg pentium is classed as a good modern machine and generally they run
either win95 or perhaps win98 se if it does not bog the machine down
too much with its extra overhead.

I suggest that win9x will be around in large quantities for some years
to come as the later versions do not offer any great advantage in
functionality terms. NT4, win2k and later will be used by the corporate
sector as they become available but it is a mistake to assume that
the rest of the world can afford to follow.

Regards,

[email protected]

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

[Borje Hagsten]

One should always try to keep up with latest technology, yes, but also
try to reason with intelligense. MS is not doing the later, if they
abandon majority and only target corporate users. But they probably
have a long term plan for it..

One interesting aspect is that a year ago, the IT hype was at its peek,
with companies building fortunes on promises. Today time has come to
fulfill their high promises and one by one goes down because dreams and
real life situations are not the same. Sweden is good example - "broadband
for all" was the promise and even the government poored enormous amounts
of taxpayer's money into the IT sector.

Today, most of those companies has gone bankrupt and not many broadband
projects are still alive. Several of my friends have lost their jobs in
that sector this spring and not many of the people I talk to has got what
they were promised to get long time ago, a working broadband connection.

This is why I have a gut feeling world economy is in for a rough ride
later on this year, when people discover MS' new system is not for
majority of current users. Maybe history is about to repeat itself.
Remember what happened to IBM some 15 years ago. Same thing may very
well happen to MS - when things start go wrong, it goes fast. Maybe
time finally is here for Linux.. so please - where is PB for LINUX?


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

[Eric Pearson]

Hutch --

I thought exactly the same thing a few years ago.

Meet you back here in five years and we'll see who's right.

-- Eric


------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>



[This message has been edited by Eric Pearson (edited May 13, 2001).]

[Eric Pearson]

One more point...

> I deal with programmers around the world
> and in many places, a 133 meg pentium is
> classed as a good modern machine and
> generally they run either win95 or perhaps
> win98 se if it does not bog the machine down
> too much with its extra overhead.

I agree, that's what most of the world is using. But that reinforces my point, rather than refuting it.

Just five years ago, those things were the state of the art! And today, it's an average system around the world.

Where will we be in five years?

-- Eric



------------------
Perfect Sync Development Tools
Perfect Sync Web Site
Contact Us: mailto:[email protected][email protected]</A>

[Borje Hagsten]

> Where will we be in five years?

Gone fishing..


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

[Steve Hutchesson]

I seem to agree with Borje in many things, going fishing appeals
to me. In 5 years things will have changed a lot but its by no
means predictable which way the change will go.

In one direction the technology will be different and some of the
newer ideas may be in production, memory technology has the promise
of being many powers larger than current. We may see the end of
Von Neuman logic in processing and the advent of "SMART" hardware,
not just fast number crunchers.

The other direction is not so rosy, cost and economic seperation
has the potential to make the hardware too expensive for most people
as the "Bill Gates" idea of a computer in every household slowly
dies a natural death. As operating systems go in the direction of
"web tv", the idea of a general purpose computer in the household
that can be used for a wide range of things may eventually disappear.

I personally came back to writing software to write flat memory model
software as I had known of it since about 1990 in assembler. It has
been a viable exercise and has earnt me a living but the corporate takeover
of the software industry and the effective end of open system architecture
is not something that has appealed to me.

If the new stuff is genuinely smart and has the potential to do new and
clever things, I may make another comeback but going fishing keeps
sounding like a better option than the learning curve to absorb
another set of badly documented APIs in an environment that is
controlled to corporate advantage.

Regards,

[email protected]

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