MacTech 1 to 12/MacTech-vol-1-12.toast/Reference/the cmsp digests ('94-'97)/csmp digest Vol 3 No 140

Text File | 1996-03-23 | 84KB | 2,091 lines

C.S.M.P. Digest             Mon, 11 Mar 96       Volume 3 : Issue 140

Today's Topics:

        "Apple's 3-D interface guidelines" ??
        'sysz' resources
        (Q) How to find local hard disk volumes only
        AppleEvents to-from an extension?
        Control panel code?
        DDP Listener Broadcast Problem
        Is C++ OK?  Make yourself heard!
        Removing a Gestalt entry - possible?



The Comp.Sys.Mac.Programmer Digest is moderated by Francois Pottier
([e-mail]).

The digest is a collection of article threads from the internet
newsgroups comp.sys.mac.programmer.help, csmp.tools, csmp.misc and
csmp.games. It is designed for people who read news semi-regularly and
want an archive of the discussions.  If you don't know what a
newsgroup is, you probably don't have access to it. Ask your systems
administrator(s) for details. If you don't have access to news, you
may still be able to post messages to the group by using a mail server
like anon.penet.fi (mail [e-mail] for more information).

Each issue of the digest contains one or more sets of articles (called
threads), with each set corresponding to a 'discussion' of a particular
subject.  The articles are not edited; all articles included in this digest
are in their original posted form (as received by our news server at
nef.ens.fr).  Article threads are not added to the digest until the last
article added to the thread is at least two weeks old (this is to ensure that
the thread is dead before adding it to the digest).  Article threads that
consist of only one message are generally not included in the digest.

The digest is officially distributed by two means, by email and ftp.

If you want to receive the digest by mail, send email to [e-mail]
with no subject and one of the following commands as body:
    help                        Sends you a summary of commands
    subscribe csmp-digest Your Name    Adds you to the mailing list
    signoff csmp-digest            Removes you from the list
Once you have subscribed, you will automatically receive each new
issue as it is created.

The official ftp info is ftp://ftp.dartmouth.edu/pub/csmp-digest.
Questions related to the ftp site should be directed to
[e-mail].

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

>From Rich <[e-mail]>
Subject: "Apple's 3-D interface guidelines" ??
Date: Mon, 19 Feb 1996 00:47:38 GMT
Organization: The World @ Software Tool & Die

I recently read a review of a 3D modeling product where the
reviewer recommended that the product's company "take to
heart Apple's 3-D interface guidelines"...

Does anyone know what particular document the reviewer was
talking about, and how I can get my hands on it if it exists?

Thanks,
Rich Wagner


P.S.  No, I don't work for that product's company.

+++++++++++++++++++++++++++

>From [e-mail] (matthew malevloent)
Date: Sun, 18 Feb 1996 20:27:23 -0800
Organization: CTS Network Services

In article <[e-mail]>, Rich <[e-mail]> wrote:

> I recently read a review of a 3D modeling product where the
> reviewer recommended that the product's company "take to
> heart Apple's 3-D interface guidelines"...
>
> Does anyone know what particular document the reviewer was
> talking about, and how I can get my hands on it if it exists?
>
> Thanks,
> Rich Wagner
>
>
> P.S.  No, I don't work for that product's company.

I would try the Apple Programmers Developers Association,
[e-mail] [I think].

Matthew Malevolent

+++++++++++++++++++++++++++

>From [e-mail] (Pete Gontier)
Date: Mon, 19 Feb 1996 13:00:32 -0800
Organization: Apple Computer, Inc.

In article <[e-mail]>,
Rich <[e-mail]> wrote:

 > I recently read a review of a 3D modeling product where the
 > reviewer recommended that the product's company "take to
 > heart Apple's 3-D interface guidelines"...
 >
 > Does anyone know what particular document the reviewer was
 > talking about, and how I can get my hands on it if it exists?

In general...

   <http://dev.info.apple.com/insidemac/quickdraw3d>

...and in particular...

   <http://dev.info.apple.com/insidemac/quickdraw3d/Intro-2.html#HEADING2-17>

- -

  Pete Gontier, Integer Poet, Apple Macintosh Developer Technical Support

  work      mail  <mailto:[e-mail]>
  personal  mail  <mailto:[e-mail]>
  personal  web   <http://www.ccnet.com/~gurgle>
  work      web   <http://dev.info.apple.com/dts.html>

+++++++++++++++++++++++++++

>From S Purcell <[e-mail]>
Date: 20 Feb 1996 01:23:22 GMT
Organization: Soliton Software

In article <[e-mail]>, Rich <[e-mail]> wrote:

> I recently read a review of a 3D modeling product where the
> reviewer recommended that the product's company "take to
> heart Apple's 3-D interface guidelines"...
>
> Does anyone know what particular document the reviewer was
> talking about, and how I can get my hands on it if it exists?

I believe they're referring to the article "Working in the 3rd Dimension"
in issue 15 of develop. You can get it at

"http://www.info.apple.com/cgi-bin/lister-pl?Apple.Support.Area/Developer_Servi
ces/Periodicals/develop/develop15/develop_Issue_15_"

it's in the Adobe portable document format. If you don't have the reader, you
can get it at

"http://www.adobe.com/Software/Acrobat/mac.html"

Seth Purcell, Mac C++ Programmer for Soliton Software



+++++++++++++++++++++++++++

>From [e-mail] (Uli Rueckert)
Date: 21 Feb 1996 20:48:48 GMT
Organization: private

Hi!

In article <[e-mail]>, Rich <[e-mail]> wrote:
> I recently read a review of a 3D modeling product where the
> reviewer recommended that the product's company "take to
> heart Apple's 3-D interface guidelines"...
>
> Does anyone know what particular document the reviewer was
> talking about, and how I can get my hands on it if it exists?
A preliminary version of the "QD 3D UI Guidelines" is on the cd, which is
contained in Apple's official QD3D documentation "3D Graphics Programming
With QD3D".
Maybe you'll find the guidelines on the QD3D SDK, which is available at
<ftp://ftp.info.apple.com/Apple.Support.Area/QuickDraw3D/>.

cu,
    U. Rueckert

+++++++++++++++++++++++++++

>From Rich <[e-mail]>
Date: Sat, 24 Feb 1996 18:21:49 GMT
Organization: The World @ Software Tool & Die

Uli Rueckert wrote:
>
> Hi!
>
> A preliminary version of the "QD 3D UI Guidelines" is on the cd, which is
> contained in Apple's official QD3D documentation "3D Graphics Programming
> With QD3D".
> Maybe you'll find the guidelines on the QD3D SDK, which is available at
> <ftp://ftp.info.apple.com/Apple.Support.Area/QuickDraw3D/>.
>
> cu,
>     U. Rueckert


I just wanted to thank all those who responded.  It turns out that
Uli Rueckert's response gave me what I was looking for: I actually
already had the QD-3D CD-ROM, but I hadn't found the article until
it was pointed out to me.  (Those CDs sure to pack a lot...)

The other responses were also useful, though.  For example, though the
"Working in the Third Dimension" article wasn't what I was initially
looking for, I'm glad to have gotten that (it was Email-ed to me!  Talk
about service  :-)

Thanks very much,
Rich Wagner



               It's no big thing, It's a small thing:
                         What people think

                               -- Byrne & Eno

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

>From John Moe <[e-mail]>
Subject: 'sysz' resources
Date: Thu, 22 Feb 1996 08:29:41 -0600
Organization: Sundstrand Aerospace

what is the 'sysz' resource?  I found it in several INIT's.
Is this related to disabling the shift key no extensions thing?
--John

+++++++++++++++++++++++++++

>From [e-mail] (David Rehring)
Date: Thu, 22 Feb 1996 15:29:11 -0700
Organization: GDT Softworks, Inc.

In article <[e-mail]>, John Moe
<[e-mail]> wrote:

> what is the 'sysz' resource?  I found it in several INIT's.
> Is this related to disabling the shift key no extensions thing?
> --John

John,

No.  It is used by the system to expand the system heap by whatever amount
is in the first 4 bytes of the 'sysz' resource.  For example, if your INIT
loads a 10K code resource and allocates a 50K buffer in the System Heap
(via NewPtrSys or something like it), then you would set the 'sysz'
resource to something like 65000 (in hex).

--
David Rehring
Software Engineer
GDT Softworks, Inc.
And all around wacky guy!

+++++++++++++++++++++++++++

>From [e-mail] (Patrick Stadelmann)
Date: Fri, 23 Feb 1996 10:16:40 +0100
Organization: University of Neuchatel

In article <[e-mail]>, John Moe
<[e-mail]> wrote:

> what is the 'sysz' resource?  I found it in several INIT's.
> Is this related to disabling the shift key no extensions thing?

No. 'sysz' is for "system zone". This resource specify the amount
of system heap (memory) that the INIT needs.

Patrick

--
Patrick Stadelmann <[e-mail]>

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

>From [e-mail] (Ken Victor)
Subject: (Q) How to find local hard disk volumes only
Date: Tue, 20 Feb 1996 15:29:05 -0800
Organization: Victor Consultin

using PBHGetVInfoSync i am able to find all mounted disks and i can
eliminate cd-roms (as being hardware protected - bit 7 of ioVAtrb),
foreign volumes (non zero ioVFSID), and ejected/dismounted volumes
(ioVDrvInfo <= 0).

i would also like to eliminate all floppies and ram disks. is there an
accepted way to do this?  i could eliminate floppies based on size less
than 1.5MB, but is this the best way?  i'm not sure how to eliminate ram
disks (although at least with apple's system 7.5.2 ram disk, bit 10 of
ioVAtrb seems to be set...)

thanx,
ken victor

ps. if possible, please respond via email:
[e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Marshall Clow)
Date: Tue, 20 Feb 1996 21:48:40 -0800
Organization: Aladdin Systems

In article <[e-mail]>,
[e-mail] (Ken Victor) wrote:

>using PBHGetVInfoSync i am able to find all mounted disks and i can
>eliminate cd-roms (as being hardware protected - bit 7 of ioVAtrb),
>foreign volumes (non zero ioVFSID), and ejected/dismounted volumes
>(ioVDrvInfo <= 0).
>
>i would also like to eliminate all floppies and ram disks. is there an
>accepted way to do this?  i could eliminate floppies based on size less
>than 1.5MB, but is this the best way?  i'm not sure how to eliminate ram
>disks (although at least with apple's system 7.5.2 ram disk, bit 10 of
>ioVAtrb seems to be set...)
>
What do you want to do about Zip drives?
All Apple floppy drives use the ".Sony" driver.
   So does the HD 20, if you care :-)

You could check the driver reference number to see if it's a SCSI device...

What, exactly are you trying to accomplish?

--
Marshall Clow
Aladdin Systems

"They that can give up essential liberty to obtain a little temporary
 safety deserve neither liberty nor safety." -- Benjamin Franklin
 _Historical Review of Pennsylvania_, 1759

+++++++++++++++++++++++++++

>From [e-mail] (Ken Victor)
Date: Wed, 21 Feb 1996 08:53:33 -0800
Organization: Victor Consultin

In article <[e-mail]>,
[e-mail] (Marshall Clow) wrote:

>In article <[e-mail]>,
>[e-mail] (Ken Victor) wrote:
>
>>using PBHGetVInfoSync i am able to find all mounted disks and i can
>>eliminate cd-roms (as being hardware protected - bit 7 of ioVAtrb),
>>foreign volumes (non zero ioVFSID), and ejected/dismounted volumes
>>(ioVDrvInfo <= 0).
>>
>>i would also like to eliminate all floppies and ram disks. is there an
>>accepted way to do this?  i could eliminate floppies based on size less
>>than 1.5MB, but is this the best way?  i'm not sure how to eliminate ram
>>disks (although at least with apple's system 7.5.2 ram disk, bit 10 of
>>ioVAtrb seems to be set...)
>>
>What do you want to do about Zip drives?
>All Apple floppy drives use the ".Sony" driver.
>   So does the HD 20, if you care :-)
>
>You could check the driver reference number to see if it's a SCSI device...
>
>What, exactly are you trying to accomplish?

thanx for the reply.  i'm looking for all files of certain types on local
hard disks  (sorry i can't say anymore at this time).

ken

+++++++++++++++++++++++++++

>From Jeff Pritchard <[e-mail]>
Date: Wed, 21 Feb 1996 18:50:32 -0700
Organization: NSI/INOW

Try this:

Boolean VolumeIsRamDisk(short vrefnum)
    {
    short drivenum = GetDriveNum(vrefnum);

    QHdrPtr driveQH = GetDrvQHdr();
    DrvQElPtr driveQ = DrvQElPtr(driveQH->qHead);
    DCtlHandle theDCE;

    while(driveQ)
        {
        if(driveQ->dQDrive == drivenum)
            {
            short refNum = driveQ->dQRefNum;
            theDCE = GetDCtlEntry(refNum);
            Ptr    theDriver = (*theDCE)->dCtlDriver;
            unsigned char *theName = (unsigned char *)((struct DRVRHeader
*)theDriver)->drvrName;
            if(EqualString(theName,"\p.EDisk",false,false))
                return true;
            else
                return false;
            }
        driveQ = DrvQElPtr(driveQ->qLink);
        }
    return false;
    }

short GetDriveNum(short vRefNum)
    {
    OSErr error;
    HParamBlockRec pb;
    unsigned char name[32];
    name[0] = 0;

    pb.volumeParam.ioCompletion = nil;
    pb.volumeParam.ioVolIndex = 0;
    pb.volumeParam.ioNamePtr = name;
    pb.volumeParam.ioVRefNum = vRefNum;
    error = PBHGetVInfoSync(&pb);
    if(error) return 0;

    return pb.volumeParam.ioVDrvInfo;        // stupid name for a drive number!
    }


BTW: I use the size to detect floppies.  If anybody creates a 1.5M
volume on a hard disk, they deserve what they get. 8^)

--
++++++++++++++++++++++++++++++++++++++++++++++++++++
+
+ Jeff Pritchard
+ Fremont, CA USA  (SF Bay Area)
+
+       Excessive moderation may lead to insufficiency.
+
++++++++++++++++++++++++++++++++++++++++++++++++++++

+++++++++++++++++++++++++++

>From [e-mail] (Pete Gontier)
Date: Fri, 23 Feb 1996 12:07:44 -0800
Organization: Apple Computer, Inc.

In article <[e-mail]>,
[e-mail] (Ken Victor) wrote:

 > i would ... like to eliminate all floppies and ram disks. is there an
 > accepted way to do this?

Not really. There are bazillions of third-party disk drivers out there,
and just about all of them have some unique property which invalidates any
one test. The best you can do is describe as many things about the volumes
you want to eliminate and write a test for each of those things.

 > i could eliminate floppies based on size less
 > than 1.5MB, but is this the best way?

This is basically what Finder does. Horrifying, eh?

- -

  Pete Gontier, Integer Poet, Apple Macintosh Developer Technical Support

  work      mail  <mailto:[e-mail]>
  personal  mail  <mailto:[e-mail]>
  personal  web   <http://www.ccnet.com/~gurgle>
  work      web   <http://dev.info.apple.com/dts.html>

+++++++++++++++++++++++++++

>From [e-mail] (Pete Gontier)
Date: Fri, 23 Feb 1996 12:08:55 -0800
Organization: Apple Computer, Inc.

In article <[e-mail]>, [e-mail] wrote:

 > if(EqualString(theName,"\p.EDisk",false,false))

This will find Apple's RAM disk driver, but there are lots of other RAM
disks out there.

- -

  Pete Gontier, Integer Poet, Apple Macintosh Developer Technical Support

  work      mail  <mailto:[e-mail]>
  personal  mail  <mailto:[e-mail]>
  personal  web   <http://www.ccnet.com/~gurgle>
  work      web   <http://dev.info.apple.com/dts.html>

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

>From [e-mail] (A. Grignon)
Subject: AppleEvents to-from an extension?
Date: 23 Feb 96 02:18:09 GMT
Organization: University of Iowa, Iowa City, IA, USA


I'd like to be able to send/receive apple events from an extension.
Sending them has proven to be no problem, but how might one
receive an AppleEvent, without having any sort of event loop?  If
I register a handler for a particular type of event, will it just
eventually get called when nobody else will accept the event regardless?

I can see a way to hack it work, by writing a jgne filter and peeking
at everything, but this doesn't seem to be a very elegant solution.
Surely this has been addressed by someone already?

Thanks for your help,
-Andy



+++++++++++++++++++++++++++

>From [e-mail] (Jon Pugh)
Date: Sun, 25 Feb 1996 07:43:31 GMT
Organization: Apple Computer

In article <[e-mail]>,
[e-mail] (A. Grignon) wrote:

>I'd like to be able to send/receive apple events from an extension.
>Sending them has proven to be no problem, but how might one
>receive an AppleEvent, without having any sort of event loop?  If
>I register a handler for a particular type of event, will it just
>eventually get called when nobody else will accept the event regardless?
>
>I can see a way to hack it work, by writing a jgne filter and peeking
>at everything, but this doesn't seem to be a very elegant solution.
>Surely this has been addressed by someone already?

CK Haun wrote a sample which should be in the snippets on the Apple web server.

http://dev.info.apple.com/source/index.html

It implements a cdev which sends AE.  I don't recall if it receives them,
but I think it does.  Knowing CK, he would do a decent job of it.

The simple solution is to write an application instead of a cdev.  The
Desktop Patterns cdev is really an app.  It's also scriptable.  It makes
the whoe thing simple.

Jon

--
What are YOU doing to oppose the Microsoft juggernaut?
         http://www.infoworkshop.com/~jonpugh/

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

>From [e-mail] (Ben Levitt)
Subject: Control panel code?
Date: Thu, 22 Feb 1996 16:23:49 -0800
Organization: Stanford University

Could someone send me, or point me towards the C/C++ source of some simple
control panels.  I would really find some examples handy, since I'm trying
to learn how to create one myself.

 Thanks!

 - Ben
[e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Matthew E. Axsom)
Date: Fri, 23 Feb 1996 15:33:19 GMT
Organization: En Technology Corp.

In article <[e-mail]>,
[e-mail] (Ben Levitt) wrote:

>Could someone send me, or point me towards the C/C++ source of some simple
>control panels.  I would really find some examples handy, since I'm trying
>to learn how to create one myself.
>
> Thanks!
>
> - Ben
>[e-mail]

Try this url...

<ftp://mirror.apple.com//mirrors/Info-Mac.Archive/dev/src/
  cw-cdev-framework-111-cpp.hqx>

-Chewey
Matthew E. Axsom                                        En Technology Corp.
[e-mail]                     Macintosh Software Craftsperson
+--------------------------------------------------------------------------+
->>(Speaking for myself and *not* En Technology, I submit the following)<<-
"When you are in it up to your ears, keep your mouth shut."

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

>From Jeff Walker <[e-mail]>
Subject: DDP Listener Broadcast Problem
Date: 21 Feb 1996 00:38:00 GMT
Organization: Ultimate Technographics Inc


I am experiencing a problem with the Multi-Node feature of AppleTalk:

        When my listener receives any BROADCAST packet from a process
running on the same machine, the packet length value is bogus.

   Upon entry to my listener code:

        ( The low word of D1 is supposed to be the length to read )
        The D1 register is 0x000D0000 ( the low word being the length )

        The D1 register is supposed to be 0x0000XXXX ( xxxx being short
                length )


        The socket listener for a MultiNode is ( to my knowledge )
exactly like a standard DDP Listener, except that it is always passed an
extended header and receives Broadcasts.

        The listener works correctly for broadcasts from other machines
and targeted packets from the same machine.

    I've read everyhing that I can get my hands on :
    Tech notes
    Inside mac networking
    inside appletalk
    and a copius amount of experimentation.

Everything is by the book. Somebody please help!

Sometimes I hate the macintosh
Jeff

+++++++++++++++++++++++++++

>From [e-mail] (Jon Hueras)
Date: Mon, 26 Feb 1996 09:12:56 -0500
Organization: The Internet Access Company

In article <4gdph8$[e-mail]>, Jeff Walker
<[e-mail]> wrote:

> I am experiencing a problem with the Multi-Node feature of AppleTalk:
>
>         When my listener receives any BROADCAST packet from a process
> running on the same machine, the packet length value is bogus.
>
>    Upon entry to my listener code:
>
>         ( The low word of D1 is supposed to be the length to read )
>         The D1 register is 0x000D0000 ( the low word being the length )
>
>         The D1 register is supposed to be 0x0000XXXX ( xxxx being short
>                                 length )
>
>
>         The socket listener for a MultiNode is ( to my knowledge )
> exactly like a standard DDP Listener, except that it is always passed an
> extended header and receives Broadcasts.
>
>         The listener works correctly for broadcasts from other machines
> and targeted packets from the same machine.
>
>         I've read everyhing that I can get my hands on :
>         Tech notes
>         Inside mac networking
>         inside appletalk
>         and a copius amount of experimentation.
>
> Everything is by the book. Somebody please help!
>
> Sometimes I hate the macintosh
> Jeff

I reported this bug to Apple almost 4 years ago. Can it really still be
unfixed? Do you happen to know what version of AppleTalk this is happening
on for you?

In any case, here's a copy of my original bug report, complete with the
code I came up with to work around the bug:

   While working on an application that makes use of AppleTalk v57's "multiple
   node architecture" I encountered what is, I believe, a bug.

   Whenever a packet is broadcast from the local machine, whether via a
   NetWrite call or via a writeDDP call from the "user node", the multinode
   receiver routine gets called with an incorrect value in D1 (always zero, I
   think). If a broadcast packet comes in from an external network, D1 always
   contains the number of bytes remaining to be received from packet, as
   expected.

   My workaround involves reconstructing the correct value for D1 using the
   packet length in the RHA and the count of the number of bytes already in
   the RHA derived from A3:

           move.l  a3, d2                             ;how many bytes in RHA?
           sub.l   a2, d2
           subq.l  #toRHA, d2
           cmp.l   #lapHdSz+ddpLength+2, d2           ;big enough?
           blo.s   FlushPkt                           ;no -->
           cmp.l   #lapHdSz+ddphSzLong+ddpMaxData, d2 ;too big?
           bhi.s   FlushPkt                           ;yes -->
           move.w  toRHA+lapHdSz+ddpLength(a2), d0    ;does d1 agree with
           and.w   #ddpLenMask, d0                    ;packet length?
           subq.w  #2, d0                             ;(Apple bug filter)
           cmp.w   d0, d1
           beq.s   LengthOK                           ;yes -->
           cmp.b   #$FF, toRHA+lapDstAdr(a2)          ;broadcast pkt?
           bne.s   FlushPkt                           ;no --> inexcusable error!
           move.w  d0, d1                             ;yes, fix up byte count

I did get an acknowedgement of this bug report from DTS as follows:

   ...just received word from AppleTalk engineering that the scenario which
   you've reported is a "likely" problem, one which they will investigate
   fixing in the next revision of AppleTalk.  The engineer commented that
   your solution is the "right" thing to do in the meantime.

Perhaps the bug was fixed and any attempt to document it swept under the
rug. If the bug persists in recent AppleTalk versions, though, that's
scary.

   Jon

+++++++++++++++++++++++++++

>From [e-mail] (Jon Hueras)
Date: Mon, 26 Feb 1996 09:51:51 -0500
Organization: The Internet Access Company

After I sent the preceding followup, I scanned a directory of all my
Developer's CD for the work Multinode and discovered a *new* addition to
Inside Mac (first appreared on the 9/95 CD) - Networking: Multinode
Architecture. Within, I discovered the following paragraphs:

  A multinode receive routine is similar in concept to a socket listener that
  receives packets addressed to a specific socket. The chapter “Datagram
  Delivery Protocol (DDP)” in this book includes a sample socket listener. To
  create a receive routine, perform the following steps:

    1. [...]

    2. Determine the number of bytes that have already been read into the .MPP
       driver’s internal buffer, called the RHA.

       To do this, subtract the beginning address of the read-header area (RHA)
       from the value in register A3, which points past the last byte read into
       the RHA. To locate the offset at the beginning of the RHA, you can use
       the toRHA equate.

Although this doesn't explicitly warn you *not* to trust D1, if you follow
their instructions you'll be doing what my workaround does.

Checking back, I see that the Tech Note "NW 13 - AppleTalk: The Rest of
the Story" says basically the same thing, so I guess there was a doc "fix"
after all. Don't know why they couldn't fix AppleTalk itself; they must
have desperately needed D1 for something else... :-)

   Jon

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

>From [e-mail] (Mark Chrisman)
Subject: Is C++ OK?  Make yourself heard!
Date: 14 Feb 1996 12:00:54 GMT
Organization: Dept. of Computer & Information Science, NCTU, Taiwan

For my game, I'm working on some modules that I intend to be re-usable
tools for other game programmers (that's you!).

I'm planning an object-oriented approach, using C++.  My question is
this:  How many of you would not consider using a tool if it's in C++?
(To use the tools, you would need to know C++, because you would have to
define some subclasses.)

Objections based on the computational overhead of OOP don't count.  Speed
is not an issue here, because the code is for things (e.g. preference
files) which won't be part of a real-time game loop.

-Mark

+++++++++++++++++++++++++++

>From [e-mail] (Jason Regier)
Date: Wed, 14 Feb 1996 09:43:24 -0800
Organization: Qualcomm, Inc.

In article <[e-mail]>,
[e-mail] (Mark Chrisman) wrote:

> How many of you would not consider using a tool if it's in C++?

If this is an application, the language in which you coded it should be
transparent to me or any other user.  I want to know that it always does
the job right, and if possible, does it in a quick and easy fashion.

But if you're thinking of supplying your own game architecture or code
snippets for other programmers, that's another story.  I'd say C++ is a
great vehicle for reusable code.  IMHO, object oriented languages allow
you to more easily maintain and upgrade an existing code base.  There is a
speed and possible memory hit from the object oriented approach, and it
can be substantial if you don't know what you're doing.  But if you code
well in C++, you should be able to emerge with some really nice code which
is also easy to upgrade and maintain.

Of course, haters of C++ will tell you this is also possible in a
procedure- based language like C or Pascal.  They'll also remind you that
C++ gives you more than enough rope to hang yourself.  But it's your call!

Jason

--
Jason Regier
GlobalStar Software Engineer
QUALCOMM, Inc.
[phone]
[e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Mark Chrisman)
Date: 14 Feb 1996 22:25:12 GMT
Organization: Dept. of Computer & Information Science, NCTU, Taiwan

In article <[e-mail]>,
[e-mail] (Jason Regier) wrote:

|In article <[e-mail]>,
|[e-mail] (Mark Chrisman) wrote:
|
|> How many of you would not consider using a tool if it's in C++?
|
|if you're thinking of supplying your own game architecture or code
|snippets for other programmers... I'd say C++ is a
|great vehicle for reusable code.  IMHO, object oriented languages allow
|you to more easily maintain and upgrade an existing code base.  There is a
|speed and possible memory hit from the object oriented approach, and it
|can be substantial if you don't know what you're doing.

While we're at it, I wonder if someone could say a few words about the
speed & memory hit from using C++.  I can see that there is a *small* (how
small?) overhead in having to look up address of virtual functions.  And I
can see that memory fragmentation might become a problem since C++ objects
are pointers, not handles.  Are there other drawbacks?  What advice would
you give to someone to wants to "know what they're doing"?

|But if you code
|well in C++, you should be able to emerge with some really nice code which
|is also easy to upgrade and maintain.
|
|Of course, haters of C++ will tell you this is also possible in a
|procedure- based language like C or Pascal.  They'll also remind you that
|C++ gives you more than enough rope to hang yourself.  But it's your call!

Well, yes, *I* am quite convinced that using C++ is a great idea.  My
concern is really this:  How many of these C++ haters are there?  Will my
neat tools go unused just because I decided to use C++?

(If they go unused because they're no good...well, that's another issue :-)

-Mark

+++++++++++++++++++++++++++

>From [e-mail] (Kevin Stone)
Date: 14 Feb 1996 21:06:37 GMT
Organization: South Dakota School of Mines and Technology

: For my game, I'm working on some modules that I intend to be re-usable
: tools for other game programmers (that's you!).
:
: I'm planning an object-oriented approach, using C++.  My question is
: this:  How many of you would not consider using a tool if it's in C++?
: (To use the tools, you would need to know C++, because you would have to
: define some subclasses.)

   For the sake of humanity!! Don't do it!  Resist the evil forces of C++!!

   No... really.  If you don't use virtual functions (much) and try to
   keep down on inheriting multiple classes, you should be fine.
   Undoubtably, you will see a small speed hit using C++ instead of C,
   but sometimes the good's outweigh the means.
   If speed isn't an issue for you, then I guess what I just said dosn't
   matter.

: Objections based on the computational overhead of OOP don't count.  Speed
: is not an issue here, because the code is for things (e.g. preference
: files) which won't be part of a real-time game loop.

   Like I said, if speed isn't an issue... it really dosn't matter.
   Personaly, I don't use C++ for several reasons (the main one is speed,
   and the lack there of).

   I guess what you really should be asking your self is... Do I NEED C++
   to write this library.  Sometimes the answer is yes... sometimes the
   answer is no.  If you think you can take advantage of things like virtual
   functions and inheritance to make the tool easier to use, then I'd say
   go for it! :)

BAS


+++++++++++++++++++++++++++

>From [e-mail] (Hung F. Fong)
Date: 15 Feb 1996 03:50:05 GMT
Organization: Princeton University

In article <4ftist$[e-mail]> [e-mail] (Kevin Stone) writes:
>   No... really.  If you don't use virtual functions (much) and try to
>   keep down on inheriting multiple classes, you should be fine.
>   Undoubtably, you will see a small speed hit using C++ instead of C,
>   but sometimes the good's outweigh the means.
>   If speed isn't an issue for you, then I guess what I just said dosn't
>   matter.
>
>: Objections based on the computational overhead of OOP don't count.  Speed
>: is not an issue here, because the code is for things (e.g. preference
>: files) which won't be part of a real-time game loop.
>
>   Like I said, if speed isn't an issue... it really dosn't matter.
>   Personaly, I don't use C++ for several reasons (the main one is speed,
>   and the lack there of).
>
>   I guess what you really should be asking your self is... Do I NEED C++
>   to write this library.  Sometimes the answer is yes... sometimes the
>   answer is no.  If you think you can take advantage of things like virtual
>   functions and inheritance to make the tool easier to use, then I'd say
>   go for it! :)
>
>BAS
>

I disagree completely that C++ is slower than C, based on two things:
- inlining in C++ allow you to do things where you normally would use
  function calls in C due laziness. (e.g. complex assignment statements)
- in the case of virtual functions it's the fastest way to implement
  "polymorphism" in practice than your usual switch statement.
  When do you use a virtual function? for example, let's say in your
  game with many different kinds of draw() routines. in C without
  exploiting function pointers it should look like:
    switch(object.type) {
    case PLAYER: /*blah blah blah*/
    case BULLET1: /*blah blah blah*/
    ...
    default: /* error */
    }
  In c++? simply object->draw(); !!!
  translated to C, this operation is basically equivalent to:
    (*(object->vfuncTable[DRAW]))(object);
  with the vfuncTable correctly reflecting the type of object
  you're playing with.

But then, in an inner loop, one can ALWAYS go back to procedural paradigm.
Therefore, I see absolutely no reason to abandon C++. While it gives
you the power to simplify complicated design structure and yet the
flexibility to write totally C efficency equivalent code in speed critical
portion of the program. And mind you, the latter is not a large part at all.

HFF

+++++++++++++++++++++++++++

>From [e-mail] (Francois Pottier)
Date: 15 Feb 1996 14:29:47 GMT
Organization: Ecole Normale Superieure, Paris, France

In article <[e-mail]>,
Mark Chrisman <[e-mail]> wrote:

>While we're at it, I wonder if someone could say a few words about the
>speed & memory hit from using C++.  I can see that there is a *small* (how
>small?) overhead in having to look up address of virtual functions.  And I
>can see that memory fragmentation might become a problem since C++ objects
>are pointers, not handles.

The cost of virtual function calls is ridiculously low, except maybe in
inner game loops. I use plenty of function calls in rather time-critical
code with no problems.

As for the cost of using pointers instead of handles, well, there is a
risk of fragmentation, but you are actually gaining a lot of speed by
doing that.

> What advice would you give to someone to wants to "know what they're
> doing"?

My advice would be to not worry too much about the language. Whatever
language you choose, you should worry about the design of your
API. Make it easy to understand, free of pitfalls, easily extendable,
easily reusable, well documented, etc. Careful design can be done in C
or C++. C++ does help you sometimes, for instance if your concepts map
easily to a class hierarchy. It can also bite you in the ass, for
instance if you think your concepts map to a class hierarchy and then
later discover that the mapping is not perfect, start hacking with
virtual base classes, etc.

Try to not over-use C++ features because they're "cool". Just use the
features which are natural for your architecture.


--
Francois Pottier
[e-mail]
[e-mail]
http://www.eleves.ens.fr:8080/home/pottier/

+++++++++++++++++++++++++++

>From [e-mail] (John W. Blackburne)
Date: Thu, 15 Feb 1996 23:01:37 +0800
Organization: Tempest

In article <[e-mail]>,
[e-mail] (Mark Chrisman) wrote:

:For my game, I'm working on some modules that I intend to be re-usable
:tools for other game programmers (that's you!).
:
:I'm planning an object-oriented approach, using C++.  My question is
:this:  How many of you would not consider using a tool if it's in C++?
:(To use the tools, you would need to know C++, because you would have to
:define some subclasses.)

At least one, me. I'm quite happy with C and have no immediate plans to
move to C++. I may end up working with it one day, but not for games
programs.

:Objections based on the computational overhead of OOP don't count.  Speed
:is not an issue here, because the code is for things (e.g. preference
:files) which won't be part of a real-time game loop.
:
:-Mark

John
--
John Blackburne,                     [e-mail]
Programmer  Asia, Inc. Online:       http://www.asia-inc.com
Technology consultant and trainer:   http://www.hk.super.net/~johnb

+++++++++++++++++++++++++++

>From [e-mail] (Bretton Wade)
Date: Thu, 15 Feb 1996 11:28:27 -0500
Organization: qualia, inc.

I also disagree about the speed issue. In general, the biggest speed
problems come from poor design, and OOP is a significantly better design
tool than sequential programming.

--
[e-mail]
http://www.qualia.com/

+++++++++++++++++++++++++++

>From [e-mail] (Brad Shapcott)
Date: 15 Feb 1996 20:04:22 GMT
Organization: Bell-Northern Research Ltd.

In article <[e-mail]>,
Bretton Wade <[e-mail]> wrote:
>I also disagree about the speed issue. In general, the biggest speed
>problems come from poor design, and OOP is a significantly better design
>tool than sequential programming.

An apocryphal story from a friend recounts a defense of direct system calls
into the Amiga OS from a group of amateur programmers, because they were
necessary for speed, despite the fact that Commodore advised against them
and would not guarantee compatability between OS releases.  (And predicatably,
the programs broke horribly across OS releases, meaning more hacks into the
code to fix compatability.)

Another programmer, presumably more professional, took the source for the
code in question and reprogrammed it using the documented OS calls and better
programming techniques.  The execution speed ended up faster.  The problem,
so it goes, lies not in the language choice, but in the programmer (in 90% of
cases).

Many beginning or ad hoc programmers develop a bad habit of using questionable
optimizations, whereas a bit of education in design, ADTs and execution
efficiency go a lot longer towards tight code that not only executes faster,
but is easier to maintain and upgrade.  I'm always suspicious when people
complain that they are being kept too far away from the machine, or don't
know why bubble sorts are bad, or don't comprehend the implications of
Turing Machine Equivalence.

C++ is a perfectly good language for such things.


Brad Shapcott

+++++++++++++++++++++++++++

>From [e-mail] (Chris S. Dillman)
Date: 15 Feb 1996 21:40:54 GMT
Organization: Cloudnet, St. Cloud MN (612)-240-8259 login as guest


CW C++ SPEED HIT.

Ok I wrote Death from above in mostly CW C++.
At first it was barley playable on 68K.
But after some intense optimizations.
I got it to be nice a playable on a 25Mhz 040
Full screen Pixel Doubled and 32 channel music.

Even plays on a 20Mhz 040 for for megs ram and ramdoubler with music on.

any now C++ is great.
Speed it is minimal Unless you implement a VECTOR LIB Then the math is
WAY way way slower.

Some guy on AOL did a comparision on that.

Any how I use lots of virtual FN in all my sprite code.
So whats the problem with useing virtual functions a lot???
Anyone???

(Im not a C++ Guru)



+++++++++++++++++++++++++++

>From [e-mail] (Andrew R. Southwick)
Date: 15 Feb 1996 22:05:18 GMT
Organization: ISSC

I vote for C++.  If the library is useful, I want it.  If your library has
a clean, mix-in architecture, I can use it straight.   A note on source vs.
linkable libraries: if your architecture and library interface isn't clean,
using it is a pain in the ass.  I can read your source code, figure out the
underlying mechanisms, completely rewrite the functionality, and integrate
it seamlessly with my existing project FASTER than trying to use a poorly-
written, confusing, buggy, and needlessly complex library.  I've got so
much of my own framework written and working already that trying to adhere
to someone else's poor idea of an OO framework is trouble I don't want.

[e-mail] (Francois Pottier) writes:
>The cost of virtual function calls is ridiculously low, except maybe in
>inner game loops. I use plenty of function calls in rather time-critical
>code with no problems.

Depends on how many times your "inner loop" is called.  186,000 function
calls per frame is devastatingly slow, whether its direct dispatch or
virtual-function.  For time-critical, tight, and insanely popular (i.e.
frequently called) inner-inner-loops, NO function calls is the best
choice.  Also check your cache size: if your inner loops are larger than
the code cache size, recode.  My 5DOF engine looped, 320 times, through
code that was just slightly larger than the 4K cache on the 040.  I
rewrote it to concentrate in different areas (best case, it blows the
cache once for each frame generated, average case was around 20 or so
with the map I was using).

Optimizing code for games is really an assembly and compiler-technology
process, and is independent of the language you are using.  To do things
fast, you need to know more about your tools than the simple facts of
what operator-overloading is, etc.

>> What advice would you give to someone to wants to "know what they're
>> doing"?

Learn about compiler technology.  Get a book on it, plus an assembly
language manual (if you don't know ASM yet).  Ask direct questions about
what (eg) "blowing the cache" is, how to find out if you're doing it, etc.
Disassembly your source.  Learn about register coloring.  Learn general
optimization techniques.

Optimization is an entirely different issue than C vs. C++; the differences
between these languages make only a minor impact on your performance numbers.

A related issue is virtual functions in C++ vs. direct dispatch in C
(this serves nicely as an example of many of the "slow" mechanisms in C++
versus their fast "counterparts" in C).  Virtual function calls *are*
slower than direct dispatch, but you don't use them in C++ as a parallel
to direct dispatch in C; they replace function pointers and switch
statements, not function calls themselves.

Summary: The "C++ is slower than C" argument is a communist plot.  Using
  C++ correctly is an easy task, but you have to think in C++, which does
  take a lot of effort to learn how to do.  Further, optimizing code is
  an almost entirely separate matter.


Andrew R Southwick          -={C++/OO Programmer}=-     [e-mail]
#include <stddisclaimer.h>                                 I speak not for IBM.
            -= Freedom by permission is a contradiction in terms =-


+++++++++++++++++++++++++++

>From [e-mail] (Brad Shapcott)
Date: 16 Feb 1996 16:31:03 GMT
Organization: Bell-Northern Research Ltd.

In article <4g0996$[e-mail]>,
Chris S. Dillman <[e-mail]> wrote:
>
>Any how I use lots of virtual FN in all my sprite code.
>So whats the problem with useing virtual functions a lot???
>Anyone???

Overhead for virtual functions is negligible.  Pushing a large parameter list
on the stack would be much more costly.  Another thing to be suspicious of:
programmers who are afraid of virtuals functions because of performance costs.


Brad Shapcott

+++++++++++++++++++++++++++

>From [e-mail] (Juri Munkki)
Date: 17 Feb 1996 18:27:10 GMT
Organization: Helsinki University of Technology

In article <4g2bg7$[e-mail]> [e-mail] (Brad Shapcott) writes:
>In article <4g0996$[e-mail]>,
>Chris S. Dillman <[e-mail]> wrote:
>>Any how I use lots of virtual FN in all my sprite code.
>>So whats the problem with useing virtual functions a lot???
>>Anyone???

>Overhead for virtual functions is negligible.  Pushing a large parameter list
>on the stack would be much more costly.  Another thing to be suspicious of:
>programmers who are afraid of virtuals functions because of performance costs.

I make all methods virtual, but I still use Think C with object extensions.
The nice thing about Think C is that it knows how to optimize monomorphic
methods...meaning that if you only have a single instance of a virtual
function, it becomes non-virtual at link time.

I don't think any real C++ compilers know how to do that, but please correct
me if I'm wrong. I would be delighted if Codewarrior supported it.

--
Juri Munkki [e-mail]        Life is easy when polygons are cheap.
http://www.iki.fi/jmunkki           Windsurfing: Faster than the wind.

+++++++++++++++++++++++++++

>From [e-mail] (Howard Berkey)
Date: 17 Feb 1996 22:00:47 GMT
Organization: Weyland-Yutani Thinking Machines Division

In article <4fvg0r$[e-mail]>, [e-mail] (Francois
Pottier) wrote:
>
> Try to not over-use C++ features because they're "cool". Just use the
> features which are natural for your architecture.

This is very good advice...

Alternatively, if someone was a strong vanilla-c programmer debating
whether or not to use C++ for a library, I'd suggest perhaps just using
C++ as a "better C" for the first project, taking advantage of the things
that can directly benefit a C programmer, like using inlined functions for
effeciency, simple classes rather than structures for data types, and
function overloading rather than varargs for simplicity in API development
when you have several ways you want to be able to call a function.  These
things are all very cool and can directly benefit a C programmer without
getting into C++'s more complex features like templates.

-H-

--
:::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
Howard Berkey, PP-ASEL, DoD #0944                   [e-mail]
http://www.sirius.com/~howard

+++++++++++++++++++++++++++

>From Thomas DeWeese <[e-mail]>
Date: Mon, 19 Feb 1996 10:16:43 -0500
Organization: Eastman Kodak

Mark Chrisman wrote:
>
> In article <[e-mail]>,
> [e-mail] (Jason Regier) wrote:
>
> |In article <[e-mail]>,
> |[e-mail] (Mark Chrisman) wrote:
> |
> |IMHO, object oriented languages allow
> |you to more easily maintain and upgrade an existing code base.  There is a
> |speed and possible memory hit from the object oriented approach, and it
> |can be substantial if you don't know what you're doing.
>
> While we're at it, I wonder if someone could say a few words about the
> speed & memory hit from using C++.

> I can see that there is a *small* (how small?) overhead in having to look up
> address of virtual functions.  And I can see that memory fragmentation might
> become a problem since C++ objects are pointers, not handles.  Are there
> other drawbacks?  What advice would you give to someone to wants to
> "know what they're doing"?

    The real things to look out for are not what you have listed.  The
most dangerous thing in C++ is that it tends to make you feel like you have
more control over what is happening than you actually have.  It is very
easy for a simple looking line of code to actually be a serious speed hit.

Some stuff to look out for:
    Small classes - It is very enticing to create small classes that
'help' you do your work (like a 2Dpoint class).  The thing to watch out
for is the hidden overhead of such classes.  Depending on what the class
contains the construction of such an object can easily balloon from 2 move
instructions to 20-30 instructions for setting up a V-table, calling the
constructor initializing the values and returning.  This can get much worse
if you have type converters which are constantly creating and destroying
temporaries.

    Inlines that don't get inlined.  While it is very nice that C++
offers the inline declairation take it with a grain of salt (a _very_large_
grain of salt).  There are numerous cases where inlines do not get inlined,
and worse yet those cases vary from compiler to compiler.  So as an example,
I use a set of macros that define get/set functions for data members (so I
can have public read, protected write).  I have extensively tested these
macros to make sure that the get inlined under the majority of circumstances
(they are pretty simple).  Well I was writting some test code and I moved
some code into inlines in the definition of the class (above my data member
definitions).  Suddenly my code ran _slower_.  I checked and the moved code
was being inlined...
eventually I noticed that the data accessed were _not_ being inlined because
the 'inline' functions were being used before the inline declairation.  The
compiler was completely free not to inline, and the program did the right
thing, just _very_ slowly.

    Class Data Members, for what ever reason code like the following
tends to be very poorly optomized:

class iter
{
  public:
    iter(char *start, char *end){ptr=start; endPtr=end;}
    iter &operator++(void) {ptr++; return *this}
    char get(void)const{return *ptr;}
    bool_t done(void)const {return ptr>=endPtr;}
  protected:
    char *ptr, *endPtr;
}

...
for (iter I(start,end);!I.done();I++)
   {...
     x = I.get();
    ... }

Many compilers feel the need to load and store I.ptr at every iteration,
as well as loading endPtr for the check.  Of lesser concern many
compilers refuse to optomize accross inline function boundries (moving
assignments around to keep the pipeline from stalling etc).

The real problem here is that if you _depend_ on the function being inlined
for speed you have to constantly be rechecking the code to see if it is still
being inlined because there are _lots_ of things that can code to suddenly
stop being inlined.

For this reason I have started to move away from relying on inlining, and
using the tried and true methods of macros, and hand expansion/machine
code generation.  I don't like it but it can be a nightmare to track down
what is being inlined and what isn't.  I don't have these problems with
these 'more traditional' methods.

If you are looking for flexability and speed is secondary C++ is wonderful.
But never underestimate the affect of an extra load/store to main memory,
in the wrong place, on program performance.

> My concern is really this:  How many of these C++ haters are there?  Will my
> neat tools go unused just because I decided to use C++?

    For stuff like preferences most of my comments are irrelivant.  If
a person is writting his program all in pascal they won't use your stuff.  If
most of the program is C they might learn enough C++ to use your stuff (depends
a lot on how hard your stuff is to use, ehh?).
--
                                Thomas DeWeese
[e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Nicolas Monnet)
Date: Mon, 19 Feb 1996 23:10:09 +0100
Organization: Ensieg/I.N.P.G.

In article (Dans l'article) <[e-mail]>, Thomas
DeWeese <[e-mail]> wrote (écrivait) :

> Mark Chrisman wrote:

>         The real things to look out for are not what you have listed.  The
> most dangerous thing in C++ is that it tends to make you feel like you have
> more control over what is happening than you actually have.  It is very
> easy for a simple looking line of code to actually be a serious speed hit.
>
> Some stuff to look out for:
>         Small classes - It is very enticing to create small classes that
> 'help' you do your work (like a 2Dpoint class).  The thing to watch out
> for is the hidden overhead of such classes.  Depending on what the class
> contains the construction of such an object can easily balloon from 2 move
> instructions to 20-30 instructions for setting up a V-table, calling the

It's easy to type "=" instead of "==", but I cannot imagine typing
"virtual" accidentally... except if one does not know what it means!

> constructor initializing the values and returning.  This can get much worse
> if you have type converters which are constantly creating and destroying
> temporaries.

One surely needs to understand perfectly object instanciation to properly
use C++. I'm not sure I do understand all the subtleties -- but, I
understand 68k ML and I'm always able to take a look at the code
produced...

 >         Inlines that don't get inlined.  While it is very nice that C++
 > offers the inline declairation take it with a grain of salt (a _very_large_
 > grain of salt).  There are numerous cases where inlines do not get inlined,
 > and worse yet those cases vary from compiler to compiler.  So as an example,
 > I use a set of macros that define get/set functions for data members (so I
 > can have public read, protected write).  I have extensively tested these
 > macros to make sure that the get inlined under the majority of circumstances
 > (they are pretty simple).  Well I was writting some test code and I moved
 > some code into inlines in the definition of the class (above my data member
 > definitions).  Suddenly my code ran _slower_.  I checked and the moved code
 > was being inlined...

   Then it might be helpful that the compiler generate a report about
inlining -- why it did'nt inline and why.
   Every compiler has its specifics; CW inlining scheme is described
rather precisely in a note, I remember. The issue is: if one uses C++ and
needs speed, then he needs to know those possible bottlenecks.

 > eventually I noticed that the data accessed were _not_ being inlined because
 > the 'inline' functions were being used before the inline declairation.  The
 > compiler was completely free not to inline, and the program did the right
 > thing, just _very_ slowly.
 >
 >         Class Data Members, for what ever reason code like the following
 > tends to be very poorly optomized:
 >
 > class iter
 > {
 >   public:
 >     iter(char *start, char *end){ptr=start; endPtr=end;}
 >     iter &operator++(void) {ptr++; return *this}
 >      char get(void)const{return *ptr;}
 >      bool_t done(void)const {return ptr>=endPtr;}
 >   protected:
 >     char *ptr, *endPtr;
 > }
 >
 > ...
 > for (iter I(start,end);!I.done();I++)
 >    {...
 >      x = I.get();
 >     ... }
 >
 > Many compilers feel the need to load and store I.ptr at every iteration,
 > as well as loading endPtr for the check.  Of lesser concern many
 > compilers refuse to optomize accross inline function boundries (moving
 > assignments around to keep the pipeline from stalling etc).
 >
 > The real problem here is that if you _depend_ on the function being inlined
 > for speed you have to constantly be rechecking the code to see if it is still
 > being inlined because there are _lots_ of things that can code to suddenly
 > stop being inlined.
 >
 > For this reason I have started to move away from relying on inlining, and
 > using the tried and true methods of macros, and hand expansion/machine
 > code generation.  I don't like it but it can be a nightmare to track down
 > what is being inlined and what isn't.  I don't have these problems with
 > these 'more traditional' methods.
 >
 > If you are looking for flexability and speed is secondary C++ is wonderful.
 > But never underestimate the affect of an extra load/store to main memory,
 > in the wrong place, on program performance.

<snip>
> --
>                                                                 Thomas DeWeese
> [e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Paul Sexton)
Date: Mon, 19 Feb 1996 23:55:23 -0500
Organization: Vnet Internet Access, Charlotte, NC - [e-mail]

In article <[e-mail]>,
[e-mail] (Jason Regier) wrote:

> In article <[e-mail]>,
> [e-mail] (Mark Chrisman) wrote:
>
> > How many of you would not consider using a tool if it's in C++?
>
The main problem is that c++ libraries tend to be compiler specific, so
you wind up supporting (at least recompiling) individually for all the
major systems.  CW has had problems with libraries breaking on each
release.

I use C++ exclusively (even on "pure" C code that doesn't need OOP) to get
the extra features and convenience.  But if I was going to put out a
library for others to use, I'd probably use straight C.

Paul

--

This posting was produced in a Microsoft free environment
for your protection.

+++++++++++++++++++++++++++

>From [e-mail] (Jason Regier)
Date: Tue, 20 Feb 1996 10:24:54 -0800
Organization: Qualcomm, Inc.

In article <[e-mail]>, Thomas DeWeese
<[e-mail]> wrote:

> Some stuff to look out for:
>         Small classes - It is very enticing to create small classes that
> 'help' you do your work (like a 2Dpoint class).  The thing to watch out
> for is the hidden overhead of such classes.
>
Yep, this is totally true.  But if you know C++ well, then you should know
when constructors and destructors are called, and when default
constructors and destructors are created, etc.  People say this all the
time-- C++ has a lot of flexibility, but it also has a lot of pitfalls as
well.

>         Inlines that don't get inlined.
>
Well, I think inlining is a pretty nice feature, supported exclusively in
C++.  But like all "features", there are some pretty serious drawbacks, as
you mentioned.  Not all compilers inline all functions declared as inlined
(especially inlined virtual functions, which very few compilers inline).
That's the breaks, I guess, but if the compiler inlines like it should,
you're still somewhat better off than you are in C.

Ah, but then why not inline all your functions?  One of the major
drawbacks to inlining is the possible resulting code bloat, which means
bigger application sizes.  And if you're trying to optimize some code and
you inline it, there's the possibility that the generated code won't fit
in the cache, and you may actually degrade the performance of your code.
It all sorta depends.  Inlining is a nice option, but it's not a panacea
to all programming problems.  Like any C++ feature, use it judiciously.

For anybody who's just listening in on this thread, don't be fooled into
thinking that the above list is an extensive list of C++ pitfalls.
There's a ton of little gotchas, and I think it's better to learn about
them from a good C++ book instead of a news thread.

Jason

--
Jason Regier
GlobalStar Software Engineer
QUALCOMM, Inc.
[phone]
[e-mail]

+++++++++++++++++++++++++++

>From [e-mail] (Brian Gray)
Date: 23 Feb 1996 19:38:04 GMT
Organization: TM Interactive Inc.

In article <4fuahd$[e-mail]>, [e-mail]
(Hung F. Fong) wrote:

> - inlining in C++ allow you to do things where you normally would use
>   function calls in C due laziness. (e.g. complex assignment statements)

  Actually, when you inline a function in C++, the compiler often has the
option of ignoring your request and refusing to inline it, treating it as
a normal function with the standard overhead thus associated.  This
happens usually when you attempt to inline a function that is longer than
a given amount determined by the compiler.  I'd say in this respect C++
offers inlining as a crutch that may deny you support at any minute.
  The more controlled alternative, which is also available in C, is using
macros.  Keep in mind though that when the compiler decides not to inline
a function so defined in the code, it usually does it because the added
memory footprint of 50 copies of a 10-line function overrides the
importance of the gained speed.  It's nice of the compiler to look out for
my well-being, but I'd rather take responsibility for my own code.

--
Brian Gray, Multimedia Manager
TM Interactive, Inc.
9696 Culver Blvd., Suite 105
Culver City, CA 90232
[phone] x124

+++++++++++++++++++++++++++

>From "Mark H. DeForest" <[e-mail]>
Date: Mon, 26 Feb 1996 13:41:37 -0800
Organization: Cyan, Inc.

Brian Gray wrote:
>
> In article <4fuahd$[e-mail]>, [e-mail]
> (Hung F. Fong) wrote:
>
> > - inlining in C++ allow you to do things where you normally would use
> >   function calls in C due laziness. (e.g. complex assignment statements)
>
>   Actually, when you inline a function in C++, the compiler often has the
> option of ignoring your request and refusing to inline it, treating it as
> a normal function with the standard overhead thus associated.  This
> happens usually when you attempt to inline a function that is longer than
> a given amount determined by the compiler.  I'd say in this respect C++
> offers inlining as a crutch that may deny you support at any minute.

On a number of C++ compilers, you can control the inline function size (I
haven't found this in CW8, though). And some C++ compilers will automatically
turn a small function into an inline function as an optimization (kind of like
the "register" keyword is just a suggestion to the compiler), so you may be
getting this benefit even if your coding in C, but using a C++ compiler.

>   The more controlled alternative, which is also available in C, is using
> macros.

The intrusive manner of macros, not really being functions, but treated as such
(like a multi-statement macro used in an if-statement) is one of the reasons why
inline functions were created.

>  Keep in mind though that when the compiler decides not to inline
> a function so defined in the code, it usually does it because the added
> memory footprint of 50 copies of a 10-line function overrides the
> importance of the gained speed.  It's nice of the compiler to look out for
> my well-being, but I'd rather take responsibility for my own code.

Most optimizing compilers allow you to control this. You can pick compiler
setting that favor speed over memory footprint. This is much easier to change
than if you had a lot of macros and allows you to use the tool to help find the
best fit for your situation rather than changing a lot of code.


mark

- - _ -----------------------------------------------------
   /              Mark H. DeForest    mailto:[e-mail]
   \_/                                http://www.cyan.com
 C Y A N

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

>From [e-mail] (John Nagle)
Subject: Removing a Gestalt entry - possible?
Date: Tue, 20 Feb 1996 03:33:23 GMT
Organization: NETCOM On-line Communication Services (408 261-4700 guest)

     Is there any way to remove a Gestalt entry?  I'd like to do
this when a driver gets unloaded.  Resetting the Gestalt entry
to nil seems like a bad idea.  Any ideas?  Thanks.

                    John Nagle

+++++++++++++++++++++++++++

>From [e-mail] (Jump Long)
Date: 20 Feb 1996 02:58:59 -0500
Organization: America Online, Inc. (1-800-827-6364)

John Nagle wrote:
>Is there any way to remove a Gestalt entry?  I'd like to do this
>when a driver gets unloaded.  Resetting the Gestalt entry to nil
>seems like a bad idea.  Any ideas?  Thanks.

The easiest way to solve this problem is to use the GestaltValue library
from Apple. It gives you three new calls (which are traps in System 7.5
and later) that make installing, changing and removing a Gestalt selector
easy and painless. The calls in the library are:

pascal OSErr NewGestaltValue (OSType selector, long newValue);
pascal OSErr ReplaceGestaltValue (OSType selector, long replacementValue);
pascal OSErr DeleteGestaltValue (OSType selector);

I've used the GestaltValue library in lots of code and it works great.

- Jim Luther

+++++++++++++++++++++++++++

>From [e-mail] (John Nagle)
Date: Tue, 20 Feb 1996 19:06:25 GMT
Organization: NETCOM On-line Communication Services (408 261-4700 guest)

[e-mail] (Jump Long) writes:
>John Nagle wrote:
>>Is there any way to remove a Gestalt entry?  I'd like to do this
>>when a driver gets unloaded.  Resetting the Gestalt entry to nil
>>seems like a bad idea.  Any ideas?  Thanks.

>The easiest way to solve this problem is to use the GestaltValue library
>from Apple. It gives you three new calls (which are traps in System 7.5
>and later) that make installing, changing and removing a Gestalt selector
>easy and painless. The calls in the library are:

>pascal OSErr NewGestaltValue (OSType selector, long newValue);
>pascal OSErr ReplaceGestaltValue (OSType selector, long replacementValue);
>pascal OSErr DeleteGestaltValue (OSType selector);

     Do you have a URL?  Apple's technote search engine is down today.
Thanks.

                    John Nagle

+++++++++++++++++++++++++++

>From [e-mail] (Thomas Reed)
Date: Wed, 21 Feb 1996 09:50:51 -0600
Organization: Washington University

In article <[e-mail]>, [e-mail] (John Nagle) wrote:

>     Is there any way to remove a Gestalt entry?  I'd like to do
>this when a driver gets unloaded.  Resetting the Gestalt entry
>to nil seems like a bad idea.  Any ideas?  Thanks.

There's no way to remove one, but you can fake it!  ;-)

If you can find a way to return a pointer to a function in your Gestalt
handler's code resource, it's easy.  You can do this several ways.  One,
install TWO Gestalts, one that returns the address of a SetVars function
(for setting some variables internal to the code resource), and another to
tell if your driver is installed.  Two, have your handler return a pointer
to a block of data containing your return value and a pointer to SetVars.
Depends on how you want it to behave.

In the code resource, define a Boolean global variable to tell you whether
or not you're installed.  Then, when your handler is called, check the
global and return an appropriate value.  To "remove" your handler, call
SetVars to change the value of the global to FALSE.

Good luck!

-Thomas

=====================================================
Thomas Reed                     Washington University
[e-mail]               Medical School
[e-mail]            Saint Louis, MO
http://medinfo.wustl.edu/~reed
- ---------------------------------------------------
Clothes make the man.  Naked people have little or no
influence on society.  -- Mark Twain
=====================================================

Opinions posted are not the opinions of Wash. U.

+++++++++++++++++++++++++++

>From [e-mail] (Dave Nebinger)
Date: Fri, 23 Feb 1996 10:34:37 -0500
Organization: KHP Services, Inc

In article <[e-mail]>, [e-mail] (John Nagle) wrote:

>      Is there any way to remove a Gestalt entry?  I'd like to do
> this when a driver gets unloaded.  Resetting the Gestalt entry
> to nil seems like a bad idea.  Any ideas?  Thanks.
>
>                                         John Nagle

How about returning the status of the driver (loaded/unloaded)
in the Gestalt selector?  Let the gestalt user worry about whether
it is loaded or not.

Dave

==========================================================
Dave Nebinger                             [e-mail]
             The Alt.Sources.Mac Archivist
     <http://www.AmbrosiaSW.com/alt.sources.mac/>
    <ftp://ftp.AmbrosiaSW.com/pub/alt.sources.mac/>

+++++++++++++++++++++++++++

>From [e-mail] (Jonas Wallden)
Date: 25 Feb 1996 13:00:05 GMT
Organization: (none)

[e-mail] (Jump Long) writes:

>The easiest way to solve this problem is to use the GestaltValue library
>from Apple. It gives you three new calls (which are traps in System 7.5
>and later) that make installing, changing and removing a Gestalt selector
>easy and painless. The calls in the library are:
>
>pascal OSErr NewGestaltValue (OSType selector, long newValue);
>pascal OSErr ReplaceGestaltValue (OSType selector, long replacementValue);
>pascal OSErr DeleteGestaltValue (OSType selector);
>
>I've used the GestaltValue library in lots of code and it works great.

GestaltValue is a great idea, but unfortunately I haven't found a way
to link PowerPC code that calls these routines. I've looked through
the CW8 libraries and the latest Developer CDs but couldn't find a
stub library where these functions are defined.

...............  .......................  ...................................
 jonas wallden       internet e-mail            world wide web homepage
   mac hacker     [e-mail]    http://www.lysator.liu.se/~jonasw

+++++++++++++++++++++++++++

>From [e-mail] (Francois Pottier)
Date: 25 Feb 1996 15:53:09 GMT
Organization: Ecole Normale Superieure, Paris, France

In article <4gpmgl$[e-mail]>,
Jonas Wallden <[e-mail]> wrote:

>GestaltValue is a great idea, but unfortunately I haven't found a way
>to link PowerPC code that calls these routines. I've looked through
>the CW8 libraries and the latest Developer CDs but couldn't find a
>stub library where these functions are defined.

GestaltValue doesn't exist on PowerPC. This library was needed because
it was rather hard to set up a Gestalt selector with its own global
storage. Now, any code fragment comes with globals so it's much easier
to write a Gestalt selector.

Agreedly they could have provided it, if only for source code
compatibility.

--
Francois Pottier
[e-mail]
[e-mail]
http://www.eleves.ens.fr:8080/home/pottier/

+++++++++++++++++++++++++++

>From [e-mail] (Jump Long)
Date: 26 Feb 1996 02:39:14 -0500
Organization: America Online, Inc. (1-800-827-6364)

John Nagle wrote:
>Do you have a URL (to GestaltValue)?  Apple's technote search
>engine is down today. Thanks.

Sure, GestaltValue is available at
<ftp://ftp.info.apple.com/Apple.Support.Area/Developer_Services/Tool_Chest
/OS_Utilities/GestaltValue.sit.hqx>.

Jonas Wallden wrote:
>GestaltValue is a great idea, but unfortunately I haven't found
>a way to link PowerPC code that calls these routines. I've
>looked through the CW8 libraries and the latest Developer CDs
>but couldn't find a stub library where these functions are
>defined.

and then Francois Pottier wrote:
>GestaltValue doesn't exist on PowerPC. This library was needed
>because it was rather hard to set up a Gestalt selector with its
>own global storage. Now, any code fragment comes with globals so
>it's much easier to write a Gestalt selector.
>
>Agreedly they could have provided it, if only for source code
>compatibility.

To use the GestaltValue 68K library from PowerPC code, you can stick the
library in a stand-alone code 68K resource, and add some wrapper glue that
loads the code resource and calls the routines. I wrote that code tonight
and I've attached an archive below in BinHex 4.0 format.

Since the three GestaltValue routines are available to non-CFM
applications if SystemSevenFiveOrLater is defined and true, my additional
code is added to your application only if GENERATINGCFM is true or
SystemSevenFiveOrLater is false. This allows you to build your programs
FAT without changing any of your source code.

The archive also includes a little CodeWarrior PPC test program just to
show that it works.

Hope this helps(tm)

- Jim Luther

(This file must be converted with BinHex 4.0)
:&&"33b"(CA0dB@ad9Q&XG@8ZFfPd!&0*9%46593K!*!%*#N!N!55fP0*9#%!!3!
!*#Pb6'&e!NF!N!-@E(3J)""38%-J4f9cG'&XG&CKE(9P!*!2I%X!N!-j!"-"D`)
A!*!$#3#3$iB!!!-!rj!%!3#Y9QqZV9Cr(!#3"Q!!N!FMS`#3#R-T!!d14f9cG'&
XG&CKE(9P,QJ!N"%iK3#3%4B!!!(A!*!$&[q3"&4&@&408&-J!3#Q)cjlV9Cr$!#
3"J)8!*!(i3!!PK`!N!D,T9&#KA6[`i4kj(0MReF,A*`[l(9R!af,r%'Je-K!UND
M&),@[M"f"h&(Gj!!mb8c3U%3!$eNXj%rNm%%0Jckb(h!FYL(C-!FmL6(Z+!,"N%
1K32N#rY3a1F9#[mYUdcNErNc3PjI"ZM9Ul8acD@cZ$VU,"f-[AS*Dj'8(r+Y$Xk
4h5i1BaV2mN1H)1IeJDDZE0BA6QDD#f[j3PjrKP$)ZX-HGa$Bdi`)!CX[ZV4p%0[
6H##%I8'I"d25G`"$Q5$NLqENLdf@icXq0ZMfq$j6L(04!emSPh1MdL#iC,X*)V-
@P`!0!"&(CA0dB@ad9Q&XG@8ZFR0bB`#3$S6q!*!4KJ!!"a`!N!-@rj!%FR0bBe*
6483"!+e@A@qY9QRa!!!(2!#3"J69!*!%6aB!N!KBZ50%$L)UHC1S'*8%bA9TT+@
j1p$@dHa[k5!+VFfpeIC+dH469DUSme&Cp8TZQ**LVT@8,2ITKKjcNT0%RaYT#6U
1BMBmm06j[Vhf`jU0U*C99AqkE1hMZ80*ecRIlPjhaY4%qBTPMVQK6aBYl6%cM[Y
A*KDYZTGU%V0RVDDDHSqrU@VZAj4h0kfJd#H,Akm[UCr!0Y4Xb&"9p8YY$Gmj%ci
+cUK9U+ERCUJ'@DT"rkS+ZCm2V5@UVkK9UDcjVD,lDhHK1cqVM9c0EcR[3DqZ9A1
(3ZIiEJePGh3fYiH3!"`UiP4%qDJL3r88HTKk6feLR[9Gb@$&cQ5rT#5U@"q9eF"
6b'kr5P8pmpVrKLUf8-r%ceq6'-I!%qiPPm8-$(H`NZk,l[p)qSI,YrcN6$#8k(1
TUZ1ebIYc"d2RX!hKVY!"fc-4NkKQRVAlJp3fh,Drab8Q+I@Yh&ZV+1PLC6e19P@
,*'D@R(VCU4fe%9B"&b6hCZLF8&Bi%([E`ECA+LF[aMa(q+JFFkdrkmS9brEHQR&
-4CEmHS%qp-N-@@JQQhVrVBRbj9YbKfKm$cNZUVM)5KJBTm-@SiD90,q@qf90fmF
r@E2V$+TBqC-c%e893A)QTPL*SNE1UT8@KaR66@,,K"23EV+e-jZ5TfBQLRRLaTe
"Q#EemT@@NZml5S+D1pMqUTLq)8fjArE82R*0SQC@#lR@A&PI)VM4YH4U%h1lXqK
ECdbeAC!!*c&T8EPIGYqAqkA3f,ilrIKJa9PAe[X`C6Z,SR-hj0i82+(UPH%$1qk
r,rGQZVE@PhXc%hjZQ,RZR(*U'r-NbTB21KRF,"C+)D[KThQ2VkGNN!"1&FTq)Sf
hKQGH9&BKR!*bHeP9pdH1T!*1G@G-aBcN*)llZl6#cjLkDQ@b)[+496Z5al(9HkE
f9!(4L@dLA#UdePYT`P3hjCi6Zf(mckPFE!IESECA5X8mUd-21lCL'rM#RD*KQE2
S%hqlI!09)2mp4kDFeBp-14XIQh,8@PGaiFL86l!B+dCQX0@H`54+PLa["EF#ZE5
+D9a-bjEpj-bHlSH[5E6ec+Kp41JK9bKpCAK@TUb+A!8dT#2F8A-rTN69[90`3US
LfamB9Tm8@0lZ$081cfV#IXcS'BHCP65HfTfq5rM0HUhUaGa2lBdXE@)LEK2FaV#
*PqDc9p@A0#ac1DcdM6ej4T8$Gm$13iF&*qXefLmfSfZrYMXCFG3*qcXRV1ANdEp
h3LHY`TidTNbJ@r-SHFZD0$j$2611c#0j5D$IqYXmHP#Si*DA0[CjF@qN[[AELqD
fjAlCr0DK"A9P0iXpTV6hdNbL(@$2ZJTlq1@[3U1C)FHmi9H'pBN(YYmpTED0953
UcPMXC'+Ra*6fR)f*1SrpTGcMXcI2a6hB@+fli@RrTGKLVDb*A@E0B8AZ6A+KDk8
b*%ph!'c-CQ58IXC8l'LpF#4dB&2a`Pp-UA@hR"#jK@aJ'rC-m`e[G2U[$hfpCX2
iU3e69cQ5VVYEHjbhYrTfKEiZCN9B4V6heVYEkmX%X[k6YlIqb3[*+Z5R%hI5aNm
ZEl*[)NM!D-hBPqdCqd!h!,T'k1VZ&,VqeJ9@ReIh&qT(XGqb$lacpc)6rkEUK-m
!-G%mSX#D5"5BPSD**ZE[eZl,pbS!N!30&NGPFh4KE(4@B@aeC5jbFh*M,QeKDf8
!N!N,p3#3%!(A!!!)R3#3!aErN!4849K869"6)!%!Ul&M-+e@I``!N!B#63#3"J%
4!!!bB`#3"[#58E8!!#Crd+FPlr6Fpi@a1iJlZS1F,aQ&N5Fjbah`#AP-eidbI6J
Ap+#eaCq1h#JlKb(L*Z53!-FAcXN[,1lc*K*K)*mEql`j#$-lj(9Rbi$*Piii)J,
+KM+3!+V40!5TME@U&'e+!q#`#SA#@NN*9fQL3j!!aB8a52H"S$m8m(P"KKq&F8f
K[99ESpk9"TU"bF+NT089)2JQU%1KF&AM-(AUKKN!'"#6R!U6,%3Hj"MN)%XHi%`
c'R)%5`J$IBFd9l08T4V)NAZec$hN#eZJKh5RN8XEqB,3lFeP$i"9ClB"H6Kc09B
3mJE`#NP*SJ'jir"pi8%LF`Ml8)EE3hU90Y&ULFZqmf&mYr&R5eD!RY6k2,N%!!d
54f9cG'&XG&CKE(9P4faeC5jME@&VC3#3#DHm!*!3"a`!!!V'!*!$&[q3"&4&@&4
08&-J!3#VX@-`V9Cr$!#3"J0Y!*!'!EN!!))m!*!'@%C4#UP3+NeU(H4m!%2JF3H
$S,8[M0e"h0(0Kr`3F!-2p2S!mS8KKc`qN!#"B"B*Y4eki'ECS0rMaRiB)KTNdXX
'EIeC`-*KaSGLQ30rd+G05PVU4F+k*jEr'Z4!PMZEZ0S3X4#j-5QMl`q$X$Z,$IT
bpId3kHH#0,#3!1BJkXF-j$#[J38Pb1rKFaVK4e`Q3elGC4CpGaLNqh`K`#)BmAY
pAL$5T3&cQJLi3ekHPSS-LCZIlJrkF6E)J!MiF4K!eNHU36XJe9458ZVNe-'6K9+
&8&L,P"G`a2!09cEV#bFccCG%@[jq2Uq9q&r'#UX'd4F*KEQ$2'9)4i%r&#!08F@
ASp3Khkp,0S2VM%RlIXJ(-RfN)Sm,jI32,#4lF9'BGm(G%#d&FAl5Pa1RjSFFS50
PN!#l-q)Y''eH4mMJa%BHc#%IU9ddeH@(kT'&3aBAaQ4F@@fj#'U%AVNl#%P,eem
5(i9a6Air3hHBI(Z8#24lNrKDeGAf))dGLX)"`L5&&+6j-1$B(%8hK&&fEZR)+3Z
6BUK5FjX[ZUcXHf"&SVB%e['a3EI(Y`,i,,@!0Ra"(el*I"UU#3F*p`-!N!30%NG
PFh4KE(4@B@aeC8GXG@8ZD'eKDf8!N!PXYJ#3%!LG!!!1d`#3!aErN!4849K869"
6)!%!Ul&M-+e@I``!N!B-eJ#3"J1G!!![Y!#3"KE)#dlaj2%%'F#%(Z'%i5qbNT@
GNHH%%8ijjH4@[SCRNT8I05AikrJ4IS56S`5!iJpi$5pC25m#m0(M+am3p#-@Z+l
Ne'24DifXj8@14KKj4LDF(#FSi%Gq*Fc((p"E!Ma[ZDf@Fj4A#%C#`[)FMP!EPTX
IQ"9a!3`5Q5)Se,*5#F*%bB*%BGJ$9TBj6jMK8P!-Qq4X"Yra!SiVNk'DDelc(,p
bPV&XhNifPrmN+bMBM+)Q8T95-8-X-Uj"Xk,-X8R2"H'C%B"PCJfAh'5b-MB'm6'
+*eEA*RpCR@IdLDX8'G0`KLLJ92+#TjM#6MH'IV`$6+6@E4kNT'cmM1IFc'!L&A#
M3CC)D#JG%*Up&bqqU((SMYYklVUl"#q[+1(ASeQ*ZT0pFbdCm&ma(9$q0l*K2$B
mVlhq8ZdXDp,*GPbhDHL)5+k+ZDLMSF!e[[jK"mK')a590X6j#Mr*Y@d%+ERYU@0
eUja+Uk46X#Q9QlTD)d94&Hl[VM0G'N@S,9rR!*lea[(S@IZY[RrSN!"mlq8Vpmp
pGdP[8SQkR2@4iH+F'LNZ8$3pYL@lS3++QXd&YFRbX*K6R$JPda3!KR'J&(LY%la
FHSC+*U&4[MH-,5M3Q'0LT'T$,LQb`-[DaYprPbI#-QF*rT0*0Bi&SIb3!1f3!)c
0TfL8J8l#")k#Nb$UM[SR4lhAJkYB-"Em!T9QqG`HeL8BKq'#`e[$HmT2TKYmYHF
ZjKSAcTY1IU6d@,e2["lDp341T(ab9lDVqeH9j4TFAd`N(-!dV"XH'jC-[f8D8pG
ai!q)JRKm2$U0qcm(R[dil3d2!drche"1U1Td$AcIEkcL8EIhr@Nd(&2Zi$6X4Ye
"-!SLEkm0DfGl*$jD[eaEfdY&aR5[LHGpQlCm-0edI1X$8PT!Zh5DD-0!MdE2qSj
iP8B906ZRIEdX&N2&FAiK!YCpGBmplqBmqmXSIKXHleBE[-A1S(q,U2i+e[f4H!I
XGJAp"qlpJISEhQhX0a!h9f($m(pLMqGR#BcJe%0T@m"4m&SV$Vkh6D(@6p'h%C6
bI9lEUEX0&`jr$++`jqk@LTdAM*j[QTF'P[0cF9#`j-YAdi9emdS4Cdc4Na!p)e@
*J44eSZJ4L#Eqq1LBRbQQCN")kDZ8p"c92%[-c9BdDlAGMT[&kpcSpPhRFH$H5FR
`BG1ZPIYfL&c0j!GJDmRY+m46UHK*#XfL8pF$HX#i#'ac2QbbIhI@'Q&V[VSq!`d
0'&"33b"(CA0dB@ad9Q&XG@8J8Q9KC#"0C3#3"fq*!*!3#XB!!",a!*!$&[q3"&4
&@&4dG(Kd!3#VX@-`V9CP"J!!!V3!!!B9!!!"33!!!QecjSEH!*!'kLd4!!J!20X
kR#(89VQEYB6m26[h"I3D6SI#85KB-"p!QC6#dELDN!!mS83Jd"UhTCQ5T-[Z*$f
,iUk8K,'I%Jpi)*5NFAe)bL(5kE8%@MUGaYp[Rjj5#b!(FeVI9E1kdp$4!YUP3iR
N#Beadc+TKGYG960RmS3CSfE[d"f&QY"Z'*,YQ1U82+(TNe'Sd9)pffe2V62F(Kh
L01K@EcVic4E9SB66bTFmS8PTTde&YcfeJhA,cpGQ"JPelPmi'U0hI"PRpf1lli)
LCTMT8"KLR0@Nf91G"38K(m3i#MX"mj%,CR[LMccb33kA)p[$qSakXMl)iBGpd)I
Iqi-FardIj$JHrM$REcl-qIi(1FkG2XKa([9KcT%2FhlV`jcIr6$RZ3pb"2M`IT*
r)(,,Fce,RZ-S&'S$C-UKkBiQ`LLP60E$cjHb#U*J*FJ!*U19(EQ9iFQa)mm)*80
'eLi**B``FLXjCNqQ`#Y!PM)"mL)!*jj+9NE3MekJLKkHTXImcZX))bKJj"Pj6T!
!`Ehe!25mm+Ff#h[ja(k`a+'Q9GcDY&V5QllhcV#i'2*XYSpa1Fb&q*+Gjl9(E#D
QJ+K,1m9lYdkFGK@GYcEC*LDlQ'lj@cAUKLaNBK!A8%dL$4Pr@dZ[MplY@b#N4[G
!Yba1m9ZC%GaUHAZT9pl(E5jPY&4XbT1dbGTp#@U'B%BD6BVG%kGYFL)f+)"P0CZ
GMk$qUl'Jl,VHlbE+RJ89$MeDbl90e$K[%DKXZAlNLd[YJ+03[f!I`j8NbHBi*(0
[5j9b-ZVB$ZI8TrM&'P'B+2&@faj%,h+e90['"6YkG)eB`bA+VPG#+3k`eb)SQq4
kL3P("[DjX+'16BTiVeQB5Q$K9TkcT-()!&cU3re2EP$l%A@0#YA3(G-dmAV@X3[
2DGXkdf,!X#cK!Sfh6YUlY9NFR)RMqNa0NJ8aGAa5)ESJ0N'3!+L6Fh$jbSfXE(P
rHK)Nl9DDR@p%q)ke#&YeI)(eC(UrqJbG,X!#VX+`f28Ul@pR&$&kb`Mj`(6%m@F
iG#[K+XfA(cqFcaGdGN`[Aaf9@[d!B'J&F((F"%"iQ*Q+cKVG'3S@RM(9VQR`*4$
NTQ4T'020`GRa3L2RN!#ST++JD6PX,+hM(F,*M!,PBECJdX0Lk[ED$K$'0AXdYD)
6aNEm[0AKd&e60#eZ0Lh!PS($IZRBmIdUU'KDe38RMVhlLZb0MfXql-66'*6KqCZ
QF'jEMQSQI&b)kq!Cm&ep!IdDU'ZfeHar!*!%$4&38%0(CA0dB@ad9Q&XG@8ZBe*
PB@3J6@8!N!IChJ#3%!l6!!!A)J#3!aErN!4849K869"6)!%!Ul&M-+e@I``!N!B
*`!#3"J2"!!"e,`#3"Ukk#dlaj2%%'F$-)kaENkam2AEN1H(NQ0c)88,2j#ZrG8,
@Cj(MC!D`HX"V!K!'X!*F"1#MaeGq(N%rBS%!VUjVIACV4cLK9q5iC-`*@KMKH)*
rS)&&@[@dANmkUQ6bKM+K&(ABHD(mEb+)T#C"QFQC,$Y6fSaTEXd5SY(SR%44+*N
*,ie'M*!!C,DLAq55HU9IX&eV,UAL9dN#`ffdRTaCB9I0E1hbZbPT+9C)R"PE'#X
m+PP)4diX#m99#9+MT49kf%ChG#2p`T3qa%"0hXSXk"V`0qA9!Tqm5l%3MQE-QJT
VVQA114fG6DJl15+Kmq#f$P)JQja**If+jXD5p)j-`DJ'k3M9[(caiRQX`cA6HLY
0Me'H+T(`aci[$FTGr24"-ei6j2Db[[b$mcjk#E+d!R,!0eY@'M6Q3SQ-pk,3eJ8
VpMYGm"H@D9C+j5YL*L[RH8Nr0,p[31C*'dpSN!#&9C)YZ5Kh6CS#P%M)M*@j#8'
Ndprk6C6j1XL%VePIbQXHfKi`"4L1FTj,cARNi"K1cakYhVfM6R[3(Tp0Zi21q@A
r!c4(faE$*$8A4fN+LCGCXY@1$&TebG9@dpEHVLVKkFEcC`#L1-'cXkTNp*S'EhU
p8fV9!BRd%YZh)XFq$'&38L#k`V&R40b&r6EZCh%p(GL&+ScmNqQTS8M*ZU6KT'e
Y6,JeUedEQCqNIk9V2@+@32!ed-A[dc4*%r6@R6qf8Q'V%)Vc4[c'H-(lTKPD6F"
RE9r1KSN+#MT*%d4ra+E$rZ9QbfT[YaC369F&PZQ6URYaJXlqbIhX,RF-MR,$&p$
VpNI$mI4X-+A"F0Tq&5F+pMVr6LLM2d%BPMhXGq8p'VGl`l1,L'C[H2jV'lmmVJ0
-)11Z!1V18*Gb8m'5[3f3!1"Zi@B-!&!5F"E+S!R-bL,5aS&M+L,)U-+@Q5maYjA
G19ENMFEH@5I8b*UX9Y[p(RPl8RmJS"(-`d1)@I5&e09F`,+VjkC"hcaY536V2Gk
XA#JqiR-l&f!2Hf&XE@1$Pf88TN(LfJA5pjZP'`)TLe-+@aj2(KECBV2Yi4MFPY"
F"-GfF0KBKJ-b%KMZ-fX++m-CYpZRHDRMG9CG6)9`m+3iF[6akU(DF",Q"+3UcVb
aBFRd&@QqLIV$2-CC[+hq-+N"JEK!'i[eQQl@lYBj32baLiMbjcNrR'YhMGf@KI#
IfGY4m($U2C&J+kXP*ZPV#%3mp2+&LVp[eVX'$qAp4mMpcAX!j#X(*0cRr`,#[Gm
G"rqfIX`kPr1NeASkTp2r!3#3"!d98&"$4f9cG'&XG&CKE(9P9'9cG#jM)%eP!*!
(aqX!N"!5m3!!'%F!N!-@rj!%9%9B9%0A588"!+e@CbqY9Rm3!*!'!H8!N!He!!!
1mJ#3"[df8EAm)8q3!22k3&0A0ZX,*c20KBP)SE8[M0e"h0%Gj(c*M%JSM%#r&f5
jr5%aIbN+"`L6l'P'K*+3!#r-"A%6BC)`bH",jc,6-"+,ZV00BJC3VkfPVD@H50)
N55%&D3cNJPl!Z#-q%)+BmBFb!H0$2L"9m!,5"CS"QbqkY!%e9d)$+Ie5PDS'$4X
e9NYq0FlB,j3l2MESp[JfD1mdEY5`J8UCqTYfaRkKhI!&IALYpEpMHM1TUA!r$3d
98&"$4f9cG'&XG&CKE(9P9'9cG#ke)%eP!*!(M&d!N"!A)J!!)lN!N!-@rj!%68e
38N0A588"!+aHCb'Y9Rl%!!!RcJ!!"q3!!!IT!!!$'IUi-+-!N!B(q!h!VCPGDfi
kU*A6c-e5bqZLRGIDFFU2F-,)bLNrLUFRTjHFFQSRTd2qY&r$8cclY9Df2XGh4@J
pqE8)Rj!!iq3iM6`Ma`QlhL5FF(*m3&MR%6iKr1h)FF)**mFT1Ak%%dDH05(FMV#
1--)bFLc+Mc$#AmPcE`AZbUj*Pfm[TkeX2Ik%d`L,(2F)Lm`iiB4621(!H!m)3+q
c``-3!*KG!C!!MDIU(UIrZf[[*1$r`3hc1AmrRMqfp13ErIZN@r@P9UYjSH$afKc
H4TZM8kK(c2LL#P4JX9KVc8hbrBBb*)'"j!D$a0GYAV3X%b"Z0m$Plf89'relRYT
3@EC8dp$ce"C6"L,T&8lrTR`M*11T`c1qqY@$QTT[)(Q*VU'([R9VRUkYVbQ)VHU
EdHph9bqUlL`!L'AdBYQR(Z!l`Kp!(BX"eBd!&9q5Y$Ri3X@5Car0G5`dS-*%-BL
XaG[[1FG$K5bd#J!Xj+LNLL4EE,bKTXk!T@S`jjYcFid[$QIK@#%%,S&l!eP,TUM
Z'-UM4cC0C"[*L,Q'm[VdGDcI$r'G%@*TGFYTr(j#Jd)cT5Z0pHX"L[hd#qJb4@0
a1@fmDe`r0H5!PVbTmJA[KaX)RTF835c91#"aPGfj3K#2-)PQ5#Za@UXQ#fi*RL9
ZYd-CJ)I&8P8DbC0BC@mch,$5l4+p90VM15c@UR+&!"j0jZZ9!A5YVDXG*ALH4he
CFlh+b*r1Sl-d56h[P+S5$l@4cme9"1!4Bq42#r)($lGLJ+lZF5,RH!KGT`Aj0D4
kAIj2[5B(HB69q"rbX0Sm[-da*UbNUp[Q83VJ`BmV'p'9'K[+3d0jr0b*c[*3'Br
iQk')U49ZMp4VXlNQkKZ5DC,bX(0SGr'GrlB1eZp&MaKZpK-m2-XGEC%mLGDb-kS
L$fqR3J!2'QSM9XRIZPTT,8S1pqS&0bUA1&b8pRL1#`lh&!r&!)GE8eTTL1TKX6C
&p&"UF-SReEMJP--mcMRP+)m*Tac9GF)T4hJS,cMP86hqFXURK$h[P#-m*Tcb+)f
,6MR-3aR!+5Yr10b6"A["iBje4X8!KpYNGejCD1KDF28*Kk[pdqNJN4-jrR3kB4l
R(%T8M`Q(%Z&aYJ12mVMBJF-mBScKVKH22,$`SeCiXX1%H$MXife[F*)`9QQ8KkZ
YBja(69Y(K-F8bmPeK@"a@-biUD-%%,pecRK8JeAd'4lBU4cMAM'&mR$B2"k$fEA
5EAFF9PIZ$&M4`P*&!&dR1d2)5iFpq9NHN3iFUE03jr51PpP%D!aXkBIJ&Sm@#$&
"#q+YKcC#blbp%JE`"N!,jr[N%*jpR)rcIIV4l&EFq%Q#(CKj$Gi"0kL!9((GMl`
,,I-VS2@(*H$lE$[ilZlLI"Zdl)h[,Qb(P)R%cj!!c(A6I&ch`jr"V4rI0m`&%Gq
p6`i),Yea6``$mb1rJH#kYmlNZVIJk`!rQ,S)GZ!ja1IZhG"5[![[,c%Z[BccIPM
2P*k+@C%af3QUZ5CTP&VAkDEGf93Kf0S&d@0UF0TAi4e((2),lYjYKdM4Y3@%Rhi
9iG2fNU+Cfh&4kLT+Q2#TbD83V`0GPMiSP,jNCk3HLHQ,PaR!"hK3A#Aa)8AAlBS
@p"l`3$rdNU++MfJP%(iD1J3pkT-Uf8IGV2XJjF%RXb&HcI4iM#aNTa4TN!#*@e@
Up*%D3&3@ACrB#bQ2ji3%bj1U%V15MUm,30q`#mmGbrd(1-6RiMXrk#r03k`Icf5
BfPf2qGr)"V8DFF0!IN$F+Z&jAT,lcLB2,pTK'JjP-6rE,1[S`ek!qV$mmCKrhhE
3TiAf`V,fVq(@[H"&""X"jG2h,1Jc0T2VG9E%2N"qDeQ[5hRm6LNrkN(ckc0PDih
kdhdq@J264fS!MCpS%qf#a`5+$cA!V4ZV!E,LmpH8"'VJ#P119A4e#,cA81HeHHd
ZTb#ZRQ[kk8f`5T4)qN$55"qJZiiFFU+FP9&k)BUTaDhBK$Vl5VG$-*3+`Z5%@l%
M#"!PQ#,!d)0DAZ`G@RP1K$JX%AB%AVk&'p+rE##IL@)-DAH*q*U2l!"I$-SR#"l
EG%DH)bN,8FH!e#%R("d(DDP[$h,k31UP``eJ@,mEr$iM8jLJi"2k@i(a"kdB
bk5&'*EA0iF$qbFbEiNH&%(hJP`c+-r%URG"GF[3['S0*"[e2`*[J2aZDIjN%+J#
8lB"61ZU6L&SaaX"),Nplq&FMSIqSN!$RPhdl[4QIc#"18,NEfFC$ZP[%pBCT-+H
Z[MB,-`lYX-lEbYB&d$hMlc,@33CHM436"+"V!3YBArJ+k0V)Za6,f`5i5d[5+9E
b+0b%@#(&CRd2b*Zd8``*iLbI2%HaaPX!hqMh8qb@pq%fJ2qR8Zc+qk%,i,*XLU8
@`Pf)&9%XrJRS3DbDBSkGm"$qh)2Pbmf%Ta&MHm@9VI!+BMG6,'iMR6SCf!5l*4p
`+QjJUaaY-`$lPQ%9a9BIK(f)2F"i2!-(!$,CU(jq$Nl*B*D6BU9[N`3FCH(%#k1
A6YkbkB34BVjP'11A*%mMXDim1)e-C1Y#`H@5",S`CRDe#`X&TcDd2-DH"6)RbJc
+ZVbLcD-TDlGlAD+kh1AdDZTYiM,"Ub[KH3%h9U`flh*mcp)Nd&eVQ9KSljSpNiP
*'e#b&!'ib5BklFjP(Sfjh),*NJ+,Nh)@'CI6*,$"5B"Rk!mQF46#p6[f2-"AKqZ
+34lr!3!!#X"&"hlQ-@85GQ4##5-d-U,N+,Q98$*KK"&'MK%U)bYK+cX*95DH8#p
lFS```[$8NlC+$Nq9%Ia(k)#``8TQM%`CBB3G'E!M-dBBBB5YNJNPamL!N3Pl-Q#
%N4%l`JJlFT3-Q16B%fD%N3RYb6%bSJ,h5DpENa`MafiY`Xl%%dD1%43aB!,JK99
HHG-d!)$Jf`-3Q-2h#Xi8I&[aIB6G!%M3&mLBC5Z[f8#H2i0,-&QXI8j!%!Qd3Pk
2*8QF&j(9I'pDRl*85dGUNI(D28KJJiFeJ'pIp-L*c0!'m"$MYAN%`48me0HGFG4
536A-NYCli1"`Ep)XD$!1pmM5jC8kE4&h-LJP,YEJ#9NP(kA$dL$1RY)*pLr*Bl+
%rBShIrH&m4jDN!#pM0ZYS)iJm`J[0SNT5[+BCMZUiH48Skc&mhAG'b@fhHAQcZp
&GUbk[r)!qk$AGhKpTpH([$jFe6pp`2i8R'CZE1MIdp#IaGRh(83EqS@Zb+f(Yk2
BTq"ibe[XT8epR8-P4l1QeEb'VQbZPK@kK&964d"@RH+1"&BdEhlTL-VMbG&bbG&
RY2Lf(*%D3SIBMkHMLb90XVmrMJEkV86iVDQ$*eE59i%4mNG(*$-d-MEa9mCk#JS
G`LmLMqTKC0R@9KlYqVVM"+TiCIZHF#2T19+4PA5+6N$L90m"EQ4i39#(&k#&8CI
mA+Q1h%b3!)TG0+ES"&SCpBfI)RHZ6e+U8jkL%fLMe0F,IQDiXI4ZEd-iJAC'[H%
ldJhhA21S1-&mPi!%ArMh!+E['UPNMdjNhGb&Fa"-B#iXh`r+fSaTcFH,cEfb@i"
4f56%UH&4cHEhf!p`JpVK8@9p6L[)H#RpSiBj0C!!R6qlTC04k6A"UA65bDQKNFb
BSaXrZ3fY%HkKD0e!hp$dBZ%HBVG,P&2p5mM0QE2#HHBXjKrhkdjHZMPpP96S2cU
",VlAlU4C+ZL1ETC8!dpFbfaXQ0eB`Z9h)HCj++8EfalDG!12CVc(39)bfTPDE4f
%I`M32c6cbH,d"(Elk*@UT'f(84HR6q**YSE[959Pf262Ib%K%&"33b"(CA0dB@a
d9Q&XG@8!N"3j!"-"D`)A!*!$!3!!'%F!N!F@!*!$KJ!!!`$rN!3"!+e@EkkY9Rm
F!*!'B!#3"b-c!*!+19M6B3!!:

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

End of C.S.M.P. Digest
**********************