Celestin Apprentice 2/Apprentice-Release2.iso/Information/Digests/CSMP Digest/volume 1/csmp-v1-212.txt

Text File | 1994-12-08 | 45.9 KB | 1,156 lines | [TEXT/R*ch]

C.S.M.P. Digest             Wed, 11 Nov 92       Volume 1 : Issue 212

Today's Topics:

    Regular Expressions
    Another Sound Manager Question (recording on PB)
    Deferred Tasks - What can I do?
    problem with my dialog filter
    Apple Events! How to locate a service?



The Comp.Sys.Mac.Programmer Digest is moderated by Michael A. Kelly.

The digest is a collection of article threads from the internet newsgroup
comp.sys.mac.programmer.  It is designed for people who read c.s.m.p. 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.  (This means you can't post questions to the
digest.)

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
cs.uoregon.edu).  Article threads are not added to the digest until the last
article added to the thread is at least one month 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 entire digest is available for anonymous ftp from ftp.cs.uoregon.edu
[128.223.8.8] in the directory /pub/mac/csmp-digest.  Be sure to read the
file /pub/mac/csmp-digest/README before downloading any files.  The most
recent issues are available from sumex-aim.stanford.edu [36.44.0.6] in the
directory /info-mac/digest/csmp.  If you don't have ftp capability, the sumex
archive has a mail server; send a message with the text '$MACarch help' (no
quotes) to [e-mail] for more information.

The digest is also available via email.  Just send a note saying that you
want to be on the digest mailing list to [e-mail], and you will
automatically receive each new issue as it is created.  Sorry, back issues
are not available through the mailing list.

Send administrative mail to [e-mail].


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

From: [e-mail] (Richard P. Daley)
Subject: Regular Expressions
Date: 30 Sep 92 23:19:11 GMT
Organization: Fulcrum - Salomon Brothers, NY

Does anyone know where I can find source code to implement searching through
text for regular expressions?  I'm hoping to find something like re_comp()
and re_exec() on the Sun.

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

From: [e-mail] (David Spencer)
Date: Thu, 1 Oct 1992 18:03:58 GMT
Organization: PANIX Public Access Unix, NYC

In <[e-mail]> [e-mail] (Richard P. Daley) writes:

>Does anyone know where I can find source code to implement searching through
>text for regular expressions?  I'm hoping to find something like re_comp()
>and re_exec() on the Sun.

The gnu regular expression library is at prep.ai.mit. The version in
the grep/egrep sources has been built and tested (for years) by
wizards. The separate regex library is called beta, but it's an
extended and rationalized version of the grep regex code.

It has bsd compatible calls, as well as its own comp/exec format.

The bsd regex code is on gatekeeper.dec.com.

You can also try the agrep code. It allows slightly more general
searching than extended regular expressions. It's from University of
Arizona; try archie to get the location.

dhs   [e-mail]


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

From: [e-mail] (Randal L. Schwartz)
Date: 2 Oct 92 06:36:59 GMT
Organization: Stonehenge Consulting Services; Portland, Oregon, USA

In article <[e-mail]> [e-mail] (David Spencer) writes:
   The gnu regular expression library is at prep.ai.mit. The version in
   the grep/egrep sources has been built and tested (for years) by
   wizards. The separate regex library is called beta, but it's an
   extended and rationalized version of the grep regex code.

Remember of course that this code is subject to the GNU Public
License.  And because of RMS's opinions about Apple, you can bet that
code for a Mac that uses these routines but violates the GPL would be
prosecuted swiftly and painfully.


- --
Randal L. Schwartz / Stonehenge Consulting Services [phone]
[e-mail] (guest account) [e-mail] (better for permanent record)
cute quote: "Welcome to Portland, Oregon -- home of the California Raisins!"

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

From: Reid Ellis <[e-mail]>
Organization: Alias Research, Inc., Toronto ON Canada
Date: Fri, 9 Oct 1992 19:11:33 GMT

David Spencer <[e-mail]> writes:
|The gnu regular expression library is at prep.ai.mit. ... The separate
|regex library is called beta, but it's an extended and rationalized
|version of the grep regex code.

Randal L. Schwartz <[e-mail]> writes:
|Remember of course that this code is subject to the GNU Public
|License.  And because of RMS's opinions about Apple, you can bet that
|code for a Mac that uses these routines but violates the GPL would be
|prosecuted swiftly and painfully.

Um, simply using GNU[*] code on a Mac does NOT violate the GPL.  The
political position that GNU has taken is just that -- a political
position.  There is nothing in the GPL that forbids GNU code to be
used for Macintosh products.  Period.  RMS[**] and GNU do, however, ask
that you boycott Apple's products [even to the point of developing any
code for them] because of Apple's coporate behaviour vis-a-vis look
and feel lawsuits.

So David, feel free to use and recommend the GNU regexp library to
Macintosh programmers if you don't agree with the position that has
been taken by RMS and GNU.

                        Reid

[*]  GNU = Gnu's Not Unix
[**] RMS = Richard M. Stallman [sp?]
- --
Reid Ellis
[e-mail]        ||           [e-mail]
[e-mail] ||  [phone] [work]

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

From: [e-mail] (Doug McNaught)
Date: 9 Oct 92 21:48:55
Organization: California Institute of Technology

In article <[e-mail]> Reid Ellis <[e-mail]> writes:

   David Spencer <[e-mail]> writes:
   |The gnu regular expression library is at prep.ai.mit. ... The separate
   |regex library is called beta, but it's an extended and rationalized
   |version of the grep regex code.

   Randal L. Schwartz <[e-mail]> writes:
   |Remember of course that this code is subject to the GNU Public
   |License.  And because of RMS's opinions about Apple, you can bet that
   |code for a Mac that uses these routines but violates the GPL would be
   |prosecuted swiftly and painfully.

   Um, simply using GNU[*] code on a Mac does NOT violate the GPL.  The
   political position that GNU has taken is just that -- a political
   position.  There is nothing in the GPL that forbids GNU code to be
   used for Macintosh products.  Period.  RMS[**] and GNU do, however, ask
   that you boycott Apple's products [even to the point of developing any
   code for them] because of Apple's coporate behaviour vis-a-vis look
   and feel lawsuits.

  Using GNU code on a Mac doesn't in itself violate the GPL, correct.
Randal's point is that code that does violate the GPL (as in, you
release the binary but refuse to make the source freely available)
would probably be more likely to attract a suit due to the FSF's
position.

   So David, feel free to use and recommend the GNU regexp library to
   Macintosh programmers if you don't agree with the position that has
   been taken by RMS and GNU.

  As long as source is freely available, there is no violation.
- -doug
- --
Doug McNaught              |"Sadder still to watch it die/ Then never to have
[e-mail]      | known it/ For you, the blind who once could see/
[e-mail]     | The bell tolls for thee..." --Neil Peart
  Nobody approves my opinions! Not even me, sometimes. Read at your own risk.

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

From: [e-mail] (Eric K. Olson)
Subject: Another Sound Manager Question (recording on PB)
Organization: Whitehead Institute for Biomedical Research
Date: Tue, 6 Oct 1992 18:04:58 GMT

I've been trying to get SPBRecordToFile() to do its thing for me,
on a PowerBook 170, running 7.0.1 + TuneUp 1.1.1, without too much
luck.  Here's what I'm doing (schematically):

    - Opening a data file (fsRdWrPerm).
    - Opening the default sound device with siWritePermission.
    - Getting the sample rate, sample size, number of channels,
        and compression type from SPBGetDeviceInfo().
    - Calling SetupAIFFHeader with the information received from
        SPBGetDeviceInfo(), the file refnum, and a buffer
        length (which is fixed at 64K).
    - Calling SPBRecordToFile() with the file refnum, a SPB, and
        the async flag set to FALSE.  The SPB has only
        the inRefNum and the count fields set to non-zero
        values (and the count field is set to 64K).
    - Closing the data file.
    - Closing the sound device.

I've looked at the info returned from SPBGetDeviceInfo():  it is all
very normal, 8 bits/sample, 'NONE' compression, one channel.

When I play back with SndStartFilePlay() I hear that the file
has the sound in it, but with a LOT of very gravelly noise on top of it.

Other things I've tried that work:

    - Playing sound files recorded with SndRecordToFile() with
        the _same_ playback routine as above
    - Recording and playing in memory using SPBSetupSndHeader()

I tried opening a sound channel before recording (not that it should
matter), but that didn't change anything.  I think that the playback
routine is OK, since it works with files recorded via
SndRecordToFile() (the routine that displays the recording dialog).  I
looked at the AIFF file and it seems to be in the correct format.  I
suspect that the actual samples being recorded are corrupted somehow
before being written to the disk.  Why does SndRecordToFile() work
when my routine fails?  Is the PowerBook actually unable to record
directly to disk (and is SndRecordToFile() not going directly to
disk, but faking it by going to memory first)?

I also noticed that the 'qual' and 'chac' selectors don't work in
SPBGetDeviceInfo().  Strange....

Any Ideas?

Thanks!

- -Eric


- --
Eric K. Olson
Whitehead Institute for Biomedical Research
9 Cambridge Center, Cambridge, MA     02142
[e-mail]         [phone]

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

From: [e-mail] (Eric K. Olson)
Date: 8 Oct 92 17:19:59 GMT
Organization: Whitehead Institute for Biomedical Research

In article <[e-mail]> I (Eric K. Olson) write:
>I've been trying to get SPBRecordToFile() to do its thing for me,
>on a PowerBook 170, running 7.0.1 + TuneUp 1.1.1, without too much
>luck.  Here's what I'm doing (schematically):
> [...]
>When I play back with SndStartFilePlay() I hear that the file
>has the sound in it, but with a LOT of very gravelly noise on top of it.

Well, I found the answer to this question after a lot of trial and error.
It turns out that AIFF files expect the sound data to be in two's-complement
format, so one has to turn that on (SetDeviceInfo('twos'...)) in order
to record directly to an AIFF file.  The gravelly noise is a result of
the sign bit being interpreted as a place value of 128.

Also, although the 'qual' selector doesn't work with GetDeviceInfo(),
it does work with SetDeviceInfo().  You can set it to 'good' ('MAC6' compression),
'betr' ('MAC3' compression), or 'best' ('NONE' compression).  Setting the
compression directly seems to always return an error.

Still don't know why the 'chac' selector fails.

So it would seem that you are supposed to set the 'qual' first, then
muck with other settings later.

NEW QUESTION: Can I use 11 kHz (or lower) sampling rate with 'good'
'qual' ('MAC6' compression)?

Anybody else playing with this?

- -Eric


- --
Eric K. Olson
Whitehead Institute for Biomedical Research
9 Cambridge Center, Cambridge, MA     02142
[e-mail]         [phone]

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

From: [e-mail] (Jim Reekes)
Date: 9 Oct 92 21:08:46 GMT
Organization: Apple Computer, Inc.

In article <[e-mail]>, [e-mail] (Eric
K. Olson) wrote:
>
> In article <[e-mail]> I (Eric K. Olson) write:
> >I've been trying to get SPBRecordToFile() to do its thing for me,
> >on a PowerBook 170, running 7.0.1 + TuneUp 1.1.1, without too much
> >luck.  Here's what I'm doing (schematically):
> > [...]
> >When I play back with SndStartFilePlay() I hear that the file
> >has the sound in it, but with a LOT of very gravelly noise on top of it.
>
> Well, I found the answer to this question after a lot of trial and error.
> It turns out that AIFF files expect the sound data to be in two's-complement
> format, so one has to turn that on (SetDeviceInfo('twos'...)) in order
> to record directly to an AIFF file.  The gravelly noise is a result of
> the sign bit being interpreted as a place value of 128.
>
> Also, although the 'qual' selector doesn't work with GetDeviceInfo(),
> it does work with SetDeviceInfo().  You can set it to 'good' ('MAC6' compression),
> 'betr' ('MAC3' compression), or 'best' ('NONE' compression).  Setting the
> compression directly seems to always return an error.
>
> Still don't know why the 'chac' selector fails.
>
> So it would seem that you are supposed to set the 'qual' first, then
> muck with other settings later.
>
> NEW QUESTION: Can I use 11 kHz (or lower) sampling rate with 'good'
> 'qual' ('MAC6' compression)?

Nope.  The current MACE assumes a sample rate of 22kHz.  Also, don't assume
that "good" means MACE 3:1.  It can be anyone of your sound input driver's
idea of what good means.  To them it may mean a lower sampling rate, a
different compression (i.e. Farallon 4:1), or a different filtering.

- -----------------------------------------------------------------------
Jim Reekes, Polterzeitgeist  |     Macintosh Toolbox Engineering
                             |          Sound Manager Expert
Apple Computer, Inc.         | RAll opinions expressed are mine, and do
20525 Mariani Ave. MS: 81-KS |   not necessarily represent those of my
Cupertino, CA 95014          |       employer, Apple Computer Inc.S

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

From: [e-mail] (Pete Resnick)
Subject: Deferred Tasks - What can I do?
Organization: University of Illinois at Urbana
Date: Wed, 7 Oct 1992 00:18:05 GMT

Simple question:

Can I move memory (e.g. call NewPtr) in a deferred task, i.e. one I
installed with DTInstall?

pr
- --
Pete Resnick             (...so what is a mojo, and why would one be rising?)
Graduate assistant - Philosophy Department, Gregory Hall, UIUC
System manager - Cognitive Science Group, Beckman Institute, UIUC
Internet: [e-mail]

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

From: [e-mail] (Andreas Wuertz)
Date: 7 Oct 92 09:58:30 GMT
Organization: Federal Institute of Technology

In article <[e-mail]> Pete Resnick,
[e-mail] writes:
>Simple question:
>
>Can I move memory (e.g. call NewPtr) in a deferred task, i.e. one I
>installed with DTInstall?

Simple question, simple answer: Deferred tasks are defined as tasks that
are too lengthy to execute as a direct response to an irq, so they're
deferred to execution with irqs reenabled but just BEFORE returning to
normal execution; so I'd say you CAN'T make calls to NewPtr a.s.o. But
this is only MY opinion.

Andy ([e-mail])

========================================================================
The only person who got all his work done by Friday was Robinson Crusoe.
========================================================================

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

From: [e-mail] (Ben Hekster)
Date: 7 Oct 92 10:39:50 GMT
Organization: University of Twente, Dept. of Computer Science

In article <[e-mail]> [e-mail] (Pete Resnick) writes:
>Can I move memory (e.g. call NewPtr) in a deferred task, i.e. one I
>installed with DTInstall?

I would say, no.  The only difference between a VBL and a Deferred Task
is that a VBL task executes while the processor interrupts are being
serviced, and a Deferred Task is executed after processor interrupts
are reenabled.  This is so that you don't get a System Error when
another interrupt arrives while you are still executing a (lengthy)
task.  For all other practical purposes, you are still executing at
interrupt time-- i.e., the heap may be in an inconsistent state, etc.

- --
Ben `Hackster' Hekster        | "He bides his time and thinks,
[e-mail]        |  'There must be more to life than this!'"

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

From: [e-mail] (Grobbins)
Date: 7 Oct 92 23:59:07 GMT
Organization: Apple Computer Inc., Cupertino, CA

In article <[e-mail]> [e-mail] (Pete Resnick) writes:
>Can I move memory (e.g. call NewPtr) in a deferred task, i.e. one I
>installed with DTInstall?

No, since deferred tasks must adhere to interrupt-time restrictions.
This is discussed in Jim Luther's DTM tech note (#320, for those
who still believe in numbers.)  I'll paste the relevant portion
below.

Grobbins            [e-mail]

Usual disclaimers apply.

- ---

from Tech Note 320, "Deferred Task Traps, Truths, and Tips":

How and when the Deferred Task Manager calls deferred tasks

Since understanding how something works usually answers most questions about
it, this note gives a complete explanation of how deferred tasks get called.
 I'll start with the occurrence of an interrupt.

When an interrupt occurs, the processor stops executing the current program
and control is passed to the Macintosh's primary interrupt handler.  The
primary interrupt handler saves registers A0-A3/D0-D3 and then decides what
secondary interrupt handler to dispatch to.  If the system has an MMU, was
booted 24-bit mode but is presently in 32-bit mode, then the MMU state is
saved and the MMU is switched to 24-bit mode.  If the system has power
management (i.e., is a portable Macintosh) and is in slow speed, the speed
is saved and is then switched to full speed.  Control is then passed to the
secondary interrupt handler which handles the interrupt.

When control returns to the primary interrupt handler from the secondary
interrupt handler, the primary interrupt handler checks to see if there are
any deferred tasks in the deferred task queue and checks to see if upon exit
of this interrupt, the processor priority level will be returned to 0.  If
both conditions are true, control is passed to the Deferred Task Manager to
execute those tasks.

The Deferred Task Manager checks to see if a deferred task is already
active; if so, the Deferred Task Manager exits.  Otherwise, one by one, each
deferred task in the deferred task queue is dequeued, and then called with
register A1 containing dtParm from the DeferredTask record and the processor
priority level is set to 0 (i.e., all interrupts are enabled).  When all
deferred tasks have been dequeued and called, control is returned to the
primary interrupt handler.

When control returns to primary interrupt handler from the Deferred Task
Manager, the primary interrupt handler restores the MMU state and speed if
necessary and then restores registers A0-A3/D0-D3.  Then the primary
interrupt handler returns control to the program executing before the
interrupt occurred.

With that explanation out of the way, here are a few things that may or may
not be obvious.

Deferred tasks are interrupt time code

That is, deferred tasks are run as a result of an interrupt.  A deferred
task must follow many of the same rules as a interrupt handler.  Here's a
warning to go along with the material in the Deferred Task Manager chapter
of Inside Macintosh Volume V.

Warning:    Deferred tasks are executed at the interrupt level and must
preserve all registers other than A0-A3 and D0-D3.  Your deferred task must
not make any calls to the Memory Manager, directly or indirectly, and can't
depend on handles to unlocked blocks being valid.  If it uses application
globals, it must also ensure that register A5 contains the address of the
boundary between the application globals and the application parameters; for
details, see Setting and Restoring the A5 Register in the Memory Management
chapter of Inside Macintosh Volume VI.  Your deferred task should avoid
accessing a low-memory variables or calling a trap that would access one
because while MultiFinder@ or System 7 is running, application low-memory
global variables are being swapped in and out.

Deferred tasks can be interrupted

That's the purpose of using deferred tasks - to let interrupt driven
processes keep running while your deferred task completes a lengthy task.
This means that your deferred task must be interruptible-that is, you should
not have to do anything such as timing-critical hardware access that can get
screwed up with interrupts on.

If the interrupt handler that installs the deferred task can be called again
while that deferred task is executing, then you also need to make sure the
way the interrupt handler and the deferred task code work together is
reentrant.  For example, the interrupt code should not modify a data buffer
that the deferred task is processing.

Deferred tasks are not prioritized

Deferred tasks are executed in the order they were added to the deferred
task queue no matter what interrupt level the code installing the task may
have been running at.

Deferred task latency

There may be significant latency incurred between installing a deferred task
with DTInstall and when that task is performed.  This time is not
deterministic and cannot be controlled.  Real-time response cannot be
guaranteed on Macintosh computers with or without the use of deferred tasks.

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

From: [e-mail] (Jim Reekes)
Date: 8 Oct 92 19:32:34 GMT
Organization: Apple Computer, Inc.

In article <[e-mail]>, [e-mail]
(Ben Hekster) wrote:
>
> In article <[e-mail]> [e-mail] (Pete Resnick) writes:
> >Can I move memory (e.g. call NewPtr) in a deferred task, i.e. one I
> >installed with DTInstall?
>
> I would say, no.  The only difference between a VBL and a Deferred Task
> is that a VBL task executes while the processor interrupts are being
> serviced, and a Deferred Task is executed after processor interrupts
> are reenabled.  This is so that you don't get a System Error when
> another interrupt arrives while you are still executing a (lengthy)
> task.  For all other practical purposes, you are still executing at
> interrupt time-- i.e., the heap may be in an inconsistent state, etc.

Both a VBL and a Deferred Task are considered interrupt code.  Neither
can make calls to the toolbox, allocate memory, etc.

VBL tasks are executed with all interrupts enabled.  This allows anything
to interrupt a VBL task.  The VBL manager will one run one VBL task at
any given time, so a VBL cannot interrupt a VBL.

Deferred Tasks are documented in the recent Tech Note, and it explains
the details of when and how they are executed.

Both a VBL and a Deferred Task are ran with processor interrupts enabled.

- -----------------------------------------------------------------------
Jim Reekes, Polterzeitgeist  |     Macintosh Toolbox Engineering
                             |          Sound Manager Expert
Apple Computer, Inc.         | RAll opinions expressed are mine, and do
20525 Mariani Ave. MS: 81-KS |   not necessarily represent those of my
Cupertino, CA 95014          |       employer, Apple Computer Inc.S

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

From: Jeff E Mandel <[e-mail]>
Organization: Tulane University School of Medicine
Date: Thu, 8 Oct 1992 23:10:12 GMT

In article <REEKES-081092123045@90.10.20.67> Jim Reekes,
[e-mail] writes:
>Both a VBL and a Deferred Task are considered interrupt code.  Neither
>can make calls to the toolbox, allocate memory, etc.
>

While I understand what you mean, this is not what you are saying. You
cannot call anything which isn't reentrant, or which presumes that an
unlocked block is where the handle says it is, etc., but that doesn't
mean you can't call anything in the toolbox, or even allocate memory.
There exist techniques whereby you can preallocate a private heap, for
which you write all of the HeapZone functions, which can be made to
function within the constraints of interrupt-driven code. Now, most folks
don't need this, because usually interrupt-driven code does the same
thing over and over again, and you can get by with just allocating a
suitably large number of queue elements and managing a free list (like
the nonpaged dynamic pool, for you VMS folks).  You can also do a a
number of things which lead up to screen output from interrupt-driven
code, but since it is a major pain to keep track of where things are,
it's not worth the trouble (I know, I once wrote a preemptive
dual-tasking executive which allowed me to simulataneously run a
real-time controller and the user interface, except it bothered the users
that it would draw graphs through the menus, and they'd subsequently be
sucked back up into the menubar! (the graphs, not the users)).

Since I probably planted the idea in Pete's head about Deferred Tasks,
the best thing I can suggest is to decide whether the blocks you need at
interrupt time can be of a fixed size, or not. If not, you've got to
write something that manages the memory you preallocate, and probably
guard the code with a spinlock, as it may get reenterred. Plan on long
sessions in MacsBug. If it can be of fixed size, then create two OS
queues, one which contains a list of free blocks, and the other is used
as your private event queue. It is possible to do this with only one
queue (this is how the event queue works, but it doesn't permit you as
much flexibility in priority handling).

Hope this helps,

Jeff E Mandel MD MS
Associate Professor of Anesthesiology
Tulane University School of Medicine
New Orleans, LA

[e-mail]

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

From: [e-mail] (Pete Resnick)
Date: 9 Oct 92 16:25:11 GMT
Organization: University of Illinois at Urbana

Jeff E Mandel <[e-mail]> writes:

>Since I probably planted the idea in Pete's head about Deferred Tasks,

You did. :-)

>the best thing I can suggest is to decide whether the blocks you need at
>interrupt time can be of a fixed size, or not.

They are fixed size in my case, and I will probably implement this
using an private queue. The problem was that I didn't want to
pre-allocate all of this memory since many of the blocks may never be
used, but it shouldn't be too big a deal.

Thanks to everyone for the input.

pr
- --
Pete Resnick             (...so what is a mojo, and why would one be rising?)
Graduate assistant - Philosophy Department, Gregory Hall, UIUC
System manager - Cognitive Science Group, Beckman Institute, UIUC
Internet: [e-mail]

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

From: [e-mail] (Jim Reekes)
Date: 9 Oct 92 21:05:30 GMT
Organization: Apple Computer, Inc.

In article <[e-mail]>, [e-mail]
(Jeff E Mandel) wrote:
>
> In article <REEKES-081092123045@90.10.20.67> Jim Reekes,
> [e-mail] writes:
> >Both a VBL and a Deferred Task are considered interrupt code.  Neither
> >can make calls to the toolbox, allocate memory, etc.
> >
>
> While I understand what you mean, this is not what you are saying. You
> cannot call anything which isn't reentrant, or which presumes that an
> unlocked block is where the handle says it is, etc., but that doesn't
> mean you can't call anything in the toolbox, or even allocate memory.
> There exist techniques whereby you can preallocate a private heap, for
> which you write all of the HeapZone functions, which can be made to
> function within the constraints of interrupt-driven code. Now, most folks
> don't need this, because usually interrupt-driven code does the same
> thing over and over again, and you can get by with just allocating a
> suitably large number of queue elements and managing a free list (like
> the nonpaged dynamic pool, for you VMS folks).  You can also do a a
> number of things which lead up to screen output from interrupt-driven
> code, but since it is a major pain to keep track of where things are,
> it's not worth the trouble (I know, I once wrote a preemptive
> dual-tasking executive which allowed me to simulataneously run a
> real-time controller and the user interface, except it bothered the users
> that it would draw graphs through the menus, and they'd subsequently be
> sucked back up into the menubar! (the graphs, not the users)).
>
> Since I probably planted the idea in Pete's head about Deferred Tasks,
> the best thing I can suggest is to decide whether the blocks you need at
> interrupt time can be of a fixed size, or not. If not, you've got to
> write something that manages the memory you preallocate, and probably
> guard the code with a spinlock, as it may get reenterred. Plan on long
> sessions in MacsBug. If it can be of fixed size, then create two OS
> queues, one which contains a list of free blocks, and the other is used
> as your private event queue. It is possible to do this with only one
> queue (this is how the event queue works, but it doesn't permit you as
> much flexibility in priority handling).

While you think you know what I mean, you really don't.  I was avoiding
a long discussion about all the hows and whys.  It's much more truthful to
state that you cannot call any toolbox calls.  There are few that can be,
and Apple doesn't document them because we couldn't.  The list of
exceptions
is far too great.  Besides, the movement is to avoid interrupt code
entirely.

Here's a little more data about the issue.  Not only is allocating
or moving blocks in the heap illegal during interrupt, you cannot call
routines that are non-reentrant.  This just about excludes all of the
toolbox.  The Mac toolbox uses lots of low memory globals, much more than
that Inside Mac ever talks about.  Because of this global state
information,
you cannot call a routine that depends on such a globals.

Furthermore, some of these globals are swapped during a tasks context.
For
example, FrontWindow refers to the global WindowList.  This global is
swapped for
the current process.  Even thou FrontWindow doesn't move or allocate
memory,
nor does it have a reentrancy problem, but it cannot be called a interrupt
level.

Because of this last issue, things get really sticky.  We don't have a list
of which globals are swapped for a task, and it changes for the many
releases
of the MultiFinder/Process Manager.  This is part of the reason we cannot
document entirely which calls cannot be called at interrupt level.  We do
attempt to provide a list of the ones we promise to be safe, such as
StripAddress.

- -----------------------------------------------------------------------
Jim Reekes, Polterzeitgeist  |     Macintosh Toolbox Engineering
                             |          Sound Manager Expert
Apple Computer, Inc.         | RAll opinions expressed are mine, and do
20525 Mariani Ave. MS: 81-KS |   not necessarily represent those of my
Cupertino, CA 95014          |       employer, Apple Computer Inc.S

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

From: [e-mail] (Daniel Morrow)
Subject: problem with my dialog filter
Date: 7 Oct 92 16:48:27 GMT
Organization: Drexel University

Basically, all I'm doing is checking an edit text field in my
dialog to see if there is anything in it. If not I dim out the
OK button, if so, the OK button becomes active again.

Anyway, this part works. The part that doesn't work is checking
for carriage returns when the OK button is enabled.

I do the following in my filter now:
    Set itemHit to the id of the OK button
    Set event.what to mouseDown.
    Set event.where to a point inside the OK button
    and I do call LocalToGlobal on that point.
    And lastly I set MyFilter to FALSE so ModalDialog knows
    to handle the event.

Am I missing something here?

Please help.

Thanks.
- -Dan.
[e-mail]

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

From: Ted Lowery <[e-mail]>
Organization: MacSolutions
Date: Thu, 8 Oct 1992 16:46:57 GMT

Try this...

void    FlashButton(int ButtonID);
/*  just flash the button, visual only  */
void    FlashButton(int ButtonID)
{
    Rect        itemRect;
    long        ticksLater;

    itemRect = GetDRect (ButtonID);
    InsetRect (&itemRect, 1, 1);
    InvertRoundRect (&itemRect, 8, 8);
    ticksLater = TickCount() + 10;            /* wait a sixth second...I don't know */
    while (TickCount() < ticksLater) ;        /* if that is standard, I just
guessed  */
    InvertRoundRect (&itemRect, 8, 8);
}



pascal Boolean charFilter(DialogPtr theDialog,EventRecord *theEvent,short
*itemHit);
pascal Boolean charFilter(DialogPtr theDialog,EventRecord *theEvent,short
*itemHit)
{
    Str255    theText;
    if (theEvent->what != keyDown) return(FALSE);  /* only looking for key
downs */

    switch ((theEvent->message) & charCodeMask) {
        case 0x0D:
        case 0x03:
            *itemHit = OKButton;                    /* basically tell ModalDialog we handled the
*/
            FlashButton(OKButton);            /*  keydown, but now there is an itemHit that
*/
            return(TRUE);                                         /* than ModalDialog gets to handle  */
            break;

        default:
            break;
        }
}

Hope this helps.

Ted
[e-mail]

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

From: [e-mail] (Matt Mora)
Date: 9 Oct 92 00:28:01 GMT
Organization: SRI International, Menlo Park, California

In article <[e-mail]> [e-mail] (Jon Wtte) writes:

>You should use _Delay, which is a system call that does just
>this, but also calls SystemTask. It might also do other
>things in the future. Doing only this is really bad, since
>there may be drivers that need time often.

I would hate to disagree with Jon but I don't think delay calls systemtask.
I think it just stuffs LMG "ticks" into d0 , adds the delay amount and loops
until "ticks" is bigger than d0.

I created my own Delay function that calls WNE with the sleep at 1 tick. It
works ok.


Matt







- --
___________________________________________________________
Matthew Mora                |   my Mac  [e-mail]
SRI International           |  my unix  [e-mail]
___________________________________________________________

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

Date: 9 Oct 92 09:05:55 GMT
Organization: Royal Institute of Technology, Stockholm, Sweden

In article <[e-mail]> [e-mail] (Matt Mora) writes:

   >You should use _Delay, which is a system call that does just
   >this, but also calls SystemTask. It might also do other
   >things in the future. Doing only this is really bad, since
   >there may be drivers that need time often.

   I would hate to disagree with Jon but I don't think delay calls systemtask.
   I think it just stuffs LMG "ticks" into d0 , adds the delay amount and loops
   until "ticks" is bigger than d0.

Okay, but the "might do other things in the future" still
holds :-) (Actually, it also sets the status register;
interrupt level one might think)

   I created my own Delay function that calls WNE with the sleep at 1 tick. It
   works ok.

As long as you don4t get update, or suspend, or ... events
while in the delay loop. At least you can mask those out;
AppleEvents, however, can NEVER be masked out, so you had
better be prepared handle them whenever you call WaitNextEvent.

- --
Jon W{tte, [e-mail], Sweden, Phone [phone]

"From now on, I think I'll re-label the EQ on the
 desk as Fizz and Wobble, rather than HF and LF."

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

From: [e-mail] (Andreas Wuertz)
Date: 9 Oct 92 13:34:30 GMT
Organization: Federal Institute of Technology

In article <[e-mail]> Daniel Morrow,
[e-mail] writes:
>    Set itemHit to the id of the OK button
>    Set event.what to mouseDown.
>    Set event.where to a point inside the OK button
>    and I do call LocalToGlobal on that point.
>    And lastly I set MyFilter to FALSE so ModalDialog knows
>    to handle the event.

How about setting MyFilter to TRUE, so ModalDialog knows YOU handled the
event and doesn't fiddle around with itemHit anymore?

In article <[e-mail]> Ted Lowery,
[e-mail] writes:
>void    FlashButton(int ButtonID);
>/*  just flash the button, visual only  */
>void    FlashButton(int ButtonID)
>{
>    Rect        itemRect;
>    long        ticksLater;
>
>    itemRect = GetDRect (ButtonID);
>    InsetRect (&itemRect, 1, 1);
>    InvertRoundRect (&itemRect, 8, 8);
>    ticksLater = TickCount() + 10;            /* wait a sixth second...I don't know
*/
>    while (TickCount() < ticksLater) ;        /* if that is standard, I just
>guessed  */
>    InvertRoundRect (&itemRect, 8, 8);
>}

I don't believe this! How about this one:
void    FlashButton(DialogPtr theDialog, int buttonID)
/* This will also work in System 9.5 (known as 'Violett') where buttons
   become dark red with greenish dots when the are clicked */
{
GetDItem(theDialog, buttonID, &itemType, &itemHandle, itemBox);
HiliteControl(itemHandle, 1);
Delay(10);
HiliteControl(itemHandle, 0); /* that's it! */
}

By the way:
Delay causes the system to wait for the number of ticks (sixtieths of a
second)
specified by numTicks, and returns in finalTicks the total number of
ticks from
system startup to the end of the delay.

Warning:  Don't rely on the duration of the delay being exact; it will
usually
          be accurate to within one tick, but may be off by more than
that.
          The Delay procedure enables all interrupts and checks the tick
count
          that's incremented during the vertical retrace interrupt;
however,
          it's possible for this interrupt to be disabled by other
interrupts,
          in which case the duration of the delay will not be exactly
what you
          requested.

Assembly-language note:??On exit from this procedure, register D0
contains the
                         value of the global variable Ticks as measured
at the
                         end of the delay.

Andy ([e-mail])

========================================================================
The only person who got all his work done by Friday was Robinson Crusoe.
========================================================================

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

From: [e-mail] (Jerome Chan)
Subject: Apple Events! How to locate a service?
Date: 7 Oct 92 19:33:16 GMT
Organization: The Tofu Society of Singapore


  Apple Events is a cool way to communicate between various applications.
However, I noticed a drawback. It doesn't have a services directory.
Everytime there is a need to locate a particular application
service/server, I've got to fiddle around with the PPC thingies. If the
application/server providing the service is not running, there is no way my
application can find out where the server/application is located, launch it
and continue (with the user's consent of course!).

  A services directory (sd) would be a nice thing to have. An application
would  register with the sd, indicating if it provides local and/or global
services. The sd would then inform all other sd's on the AppleTalk net of
this new addition if global services are provided. When an application
requires to connect to a particulat service, it would query the sd to find
the correct address. The sd could also keep a record of previously active
services and their locations and launch them (via AEs to the
approperiate(spelling?) finder) if requested

  Is there something else available/comming soon that would provide this
functionality? Are the software people at Apple going to provide something
like this in the next system release? Anyone know? Am I wasting my time?

  Maybe it's because it's four o'clock in the afternoon.



- ---
 The Evil Tofu

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

From: [e-mail] (Ross Brown)
Date: 8 Oct 92 15:56:41 GMT
Organization: Bell-Northern Research Ltd.

In article <[e-mail]> [e-mail] (Jerome
Chan) writes:
>
>  Apple Events is a cool way to communicate between various applications.
>However, I noticed a drawback. It doesn't have a services directory.
>Everytime there is a need to locate a particular application
>service/server, I've got to fiddle around with the PPC thingies. If the
>application/server providing the service is not running, there is no way my
>application can find out where the server/application is located, launch it
>and continue (with the user's consent of course!).

Well, there's no sure-fire programmatic way of doing it in System 7, but if
your local program knows that the remote application/server has a certain PPC
port name, or a certain NBP entity type alias assigned by PPC, and your machine
has an alias to the application/server somewhere on its disk, then everything
you describe can be done by a program on your machine without alerting the
user, with the requisite combination of IPCListPorts, StartSecureSession, and
remote Finder AEs.  This amounts to implementing a services directory on the
*client* machine rather than the *server*, which may not be what you want, but
it works.

A directory server capability of the sort you describe could be built into the
client program on the standard of signatures.  PPC allows servers to register
themselves on the network on a creator/type basis, so client programs can use
signatures to check whether a server is running.  The desktop database has a
call (PBDTGetAPPL) which can find the location of the server program given its
signature.  Once your client program mounts the remote volume where the server
program lives, you can use PBDTGetAPPL with ioVRefNum referring to the remote
volume.  The resulting value in ioAPPLParID gives you the exact directory,
which you can easily search to find the server program.  Make an alias, send a
'sope' AE to the remote Finder, and voila.

I'd like to see the result, once you get it working!

==============================================================================
Ross Brown, Dept. 7C22  < Bell-Northern Research     >  Just the facts, ma'am.
Advisor, Telemgmt Svcs  < P. O. Box 3511, Station C  >  We don't care whose
[e-mail]             < Ottawa, ON, Canada K1Y 4H7 >  opinions yours aren't.
==============================================================================

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

From: [e-mail] (Chuck Shotton)
Date: 8 Oct 1992 21:00:25 GMT
Organization: Academic Computing

In article <[e-mail]>,
[e-mail] (Jon Wtte) wrote:
>
> And still we don't have a good service dictionary in the system.
> Anyone care to try and position the HLE and the AE in the OSI
> stack? PPC should be at the same level as TCP, i.e. level 3 ?
> (starting at 0)
>

AE.... Session level.

- --_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-
Chuck Shotton                      |
Ass't Director, Academic Computing |       "This space for rent."
UT Health Science Center Houston   |
[e-mail]       |
_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-

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

Organization: Royal Institute of Technology, Stockholm, Sweden
Date: Thu, 8 Oct 1992 19:45:56 GMT

> [e-mail] (Ross Brown) writes:

> Well, there's no sure-fire programmatic way of doing it in System 7,
> but if your local program knows that the remote application/server has
> a certain PPC port name, or a certain NBP entity type alias assigned

Well, everything would be nicer if you could actually USE the
documented-but-not-implemented type/creator port format. Now
you can use the not-documented-but-I-hope-stable format <creator>ep01
(ep == Event Port ?)

> somewhere on its disk, then everything you describe can be done by a
> program on your machine without alerting the user, with the requisite
> combination of IPCListPorts, StartSecureSession, and remote Finder

Nope; you still have to use program linking to establish an
AppleEvent channel (username & password) - We work around that
using AutoGuest INIT, but that's REALLY skanky (we should, at
least, use the library version)

Does the finder listen to remote events? I thought not. Therefore
you have to write a small system extension (background application)
that listens for specific AppleEvents and relays them to the local
finder. Talk about security hole :-)

According to Really Heavy Names (i.e. the guys who did it) this
is not easily solvable, since the AE manager lies on top of the
HLE manager, which lies on top of PPC. If AE lied directly on
top of PPC, one could conceive of specific attributes for name
and password; however, the HLE manager known nothing about what
you send through it.

And still we don't have a good service dictionary in the system.
Anyone care to try and position the HLE and the AE in the OSI
stack? PPC should be at the same level as TCP, i.e. level 3 ?
(starting at 0)

Cheers,

                        / h+

- --
Jon W{tte, [e-mail], Sweden, Phone [phone]

"From now on, I think I'll re-label the EQ on the
 desk as Fizz and Wobble, rather than HF and LF."

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

From: [e-mail] (Ross Brown)
Organization: Bell-Northern Research Ltd.
Date: Fri, 9 Oct 92 20:33:08 GMT

In article <[e-mail]> [e-mail]
(Jon Wtte) writes:
>[stuff deleted]
>Does the finder listen to remote events? I thought not. Therefore
>you have to write a small system extension (background application)
>that listens for specific AppleEvents and relays them to the local
>finder. Talk about security hole :-)

Does Finder listen to remote events?  It does for me.  Check out Jon Pugh's
FinderEvents stack.  In the description of the Open Selection and Print
Selection events, he mentions that these are the only ones you can fire at a
remote Finder, and by Jove, it works.

You may have noticed that Finder always appears in the list of programs when
you are poking around the network in the PPC Browser.  In fact, there's no
Sharing dialog for Finder, so you can't tell it *not* to listen to remote
events... 'sope' and 'spri', at least.

==============================================================================
Ross Brown, Dept. 7C22  < Bell-Northern Research     >  Just the facts, ma'am.
Advisor, Telemgmt Svcs  < P. O. Box 3511, Station C  >  We don't care whose
[e-mail]             < Ottawa, ON, Canada K1Y 4H7 >  opinions yours aren't.
==============================================================================


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

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