--- In eiffel_software@yahoogroups.com, "Emmanuel Stapf [ES]"
<manus@...> wrote:
>
> > Why is this happening?
>
> Looks like something goes wrong during the precompilation phase and
since we are
> doing 3 precompilations for classic and 3 for .NET. My guess is that
the .NET
> one are failing.
>
> > On a different matter I want to replace the beta 2.0 version of the
> > Microsoft.NET Framework. Can I uninstall the the beta version of the
> > framework, install the new 2.0 framework without uninstalling
> > EiffelStudio 5.7 first?
>
> I suspect that there is something that EiffelStudio 5.7 doesn't like
in the beta
> of .NET 2.0. You can uninstall the beta and install the final
release of .NET
> 2.0 without reinstalling  EiffelStudio. Only the .NET precompiles
might need to
> be rebuilt if you need them.
>
> Regards,
> Manu
>


I uninstalled EiffelStudio 5.7, uninstalled Microsft .NET Framework
2.0 beta, installed Microsft .NET Framwork 2.0 (final non-beta
version) and reinstalled EiffelStudio 5.7  with the Borland C++
compiler selected again. This time the installation completed without
incident precompiling base, vision2 and wel.

Launching EiffelStudio and opening the Precompilation Wizard shows:

base-mt         done = no
base  done = yes
vision2-mt done = no
vision2  done = yes
wel-mt  done = no
wel  done = yes

What are "base-mt", "vision2-mt" and "wel-mt"? Were they suppose to be
precompiled during enstallation?

thanks,
Ron

> What are "base-mt", "vision2-mt" and "wel-mt"? Were they
> suppose to be precompiled during enstallation?

It is the version of base, wel and vision2 compiled in multithreaded mode. They
are only needed if you plan on developing a multithreaded application using a
precompiled library.

Regards,
Manu

Hi,

Eiffel Studio 5.7 for linux cannot display chinese
characters nor accept inputs from input method such as
scim.

Have ISE plan to work on this problem?

Cheng-Chang Wu

___________________________________________________
  ±zªº¥Í¬¡§Y®É³q ¡Ð ·¾³q¡B®T¼Ö¡B¥Í¬¡¡B¤u§@¤@¦¸·d©w¡I
  http://messenger.yahoo.com.tw/

>From: Cheng-Chang Wu <chengchangwu@...>

>Eiffel Studio 5.7 for linux cannot display chinese
>characters nor accept inputs from input method such as
>scim.

This is consistent with 5.7 compiler not accepting Unicode class texts.

>Have ISE plan to work on this problem?

Since ECMA mandates Unicode class texts (8.32.1) as an option, then they
surely have.

_________________________________________________________________
Windows Live™ Messenger has arrived. Click here to download it for free!
http://imagine-msn.com/messenger/launch80/?locale=en-gb

On Thu, 02 Nov 2006 09:32:22 +0100, Colin Adams
<colinpauladams@...> wrote:

>> From: Cheng-Chang Wu <chengchangwu@...>
>
>> Eiffel Studio 5.7 for linux cannot display chinese
>> characters nor accept inputs from input method such as
>> scim.
>
> This is consistent with 5.7 compiler not accepting Unicode class texts.
>
>> Have ISE plan to work on this problem?
>
> Since ECMA mandates Unicode class texts (8.32.1) as an option, then they
> surely have.

A couple of students have tried to create an internationalized version of
EiffelStudio including a chinese translation of the development
environment. Unfortunately, switching from STRING_8 to STRING_32
everywhere seems to create numerous problems. Still, the students have
create an internationalization framework that can be applied to Eiffel
applications. We hope to publish it soon. That way, you would first create
an English application using EiffelStudio (employing the framework) and
later create translations using the well-established .mo/.po files.

Eventually we might get a Chinese version of EiffelStudio, but that
definitely needs lots of time and work.

See http://eiffelsoftware.origo.ethz.ch/index.php/I18n

Bernd

--- Colin Adams <colinpauladams@...> »¡¡G

> >From: Cheng-Chang Wu <chengchangwu@...>
>
> >Eiffel Studio 5.7 for linux cannot display chinese
> >characters nor accept inputs from input method such
> as
> >scim.
>
> This is consistent with 5.7 compiler not accepting
> Unicode class texts.
>
> >Have ISE plan to work on this problem?
>

But not all chinese texts are encoded using Unicode.
There are at least three common chinese encoding
systems:  GB, BIG5 and Unicode. GB encoding system is
the official standard in Mainland china, but BIG5 is
de facto standard of Taiwan.

GTK can handle all of these encodings. Maybe Eiffel
Studio has masked the highest bit of characters so
they cannot be shown with GTK.

For CJK input Eiffel Studio should use GtkImContext to
get input from the user.

Cheng-Chang Wu


___________________________________________________
  ±zªº¥Í¬¡§Y®É³q ¡Ð ·¾³q¡B®T¼Ö¡B¥Í¬¡¡B¤u§@¤@¦¸·d©w¡I
  http://messenger.yahoo.com.tw/

Hi,

>
> A couple of students have tried to create an
> internationalized version of
> EiffelStudio including a chinese translation of the
> development
> environment. Unfortunately, switching from STRING_8
> to STRING_32
> everywhere seems to create numerous problems.

My problem is really not that big. I was developing
eiffel programs using Eclipse as my editor. Now I want
to  edit these programs using Eiffel Studio's editor,
but comments written in chinese can not be shown
correctely. These chinese characters are encoded with
Big5, not Unicode. Most of the GTK applications can
treat these strings as 8 bit strings and show it
according to the current locale. Only Eiffel Studio
fails to do it correctly.

The second problem is Eiffel Studio seems have no
support for XIM or gtk's immodule, so no chinese input
server can connect to Eiffel Studio.

Cheng-Chang Wu

___________________________________________________
  ±zªº¥Í¬¡§Y®É³q ¡Ð ·¾³q¡B®T¼Ö¡B¥Í¬¡¡B¤u§@¤@¦¸·d©w¡I
  http://messenger.yahoo.com.tw/

>From: Cheng-Chang Wu <chengchangwu@...>

>--- Colin Adams <colinpauladams@...> »¡¡G

>
>But not all chinese texts are encoded using Unicode.

So convert them using iconv.

_________________________________________________________________
Windows Live™ Messenger has arrived. Click here to download it for free!
http://imagine-msn.com/messenger/launch80/?locale=en-gb

>From: "Emmanuel Stapf [ES]" <manus@...>
> > When information is available, I update an EV_VERTICAL_BOX conatined
> > within a page of an EV_NOTEBOOK. The update consists in changing the
> > background color of certain widgets within the box.
>
>What kind of widgets?
>
> > I tried adding a call to refresh_now on the box, but it makes no
> > difference.
>
>You can also try to call {EV_APPLICATION}.process_events to see if it makes
>a
>difference.

It doesn't.

_________________________________________________________________
Windows Live™ Messenger has arrived. Click here to download it for free!
http://imagine-msn.com/messenger/launch80/?locale=en-gb

Manu, I have been unable to get a recurrence of this error message for the past
several days.  I had it continuously for two days; now nothing.  Obviously a
false alarm.  If it should occur again I will try to capture it precisely at
that time. Sorry for the false alarm.

Emmanuel Stapf [ES] wrote:
>
>
>  > However, I get this error message out to Stdout each time I click
>  > 'compile' or execute a compile by any means:
>  >
>  > estudio: could not launch
>  > /usr/local/Eiffel/Eiffel57/studio/spec/linux-x86/bin/ec -bench
>  >
>  > '-bench' is not [now] an available option to the ec command according
>  > to the output from 'ec -?'.
>
> Could you be clearer on when exactly you get this message? The new
> option to use
> is `ec -gui' however the undocumented `-bench' and `-from_bench' are still
> accepted for backward compatibility.
>
> I would check that in
> /usr/local/Eiffel/Eiffel57/studio/spec/linux-x86/bin/ you
> have `ec' with executable permission.
>
> Regards,
> Manu
>

Many vendors are making vmware appliances available with their software
preinstalled now that vmware player is a free download.  For example,
that approach allowed me to experiment with Mono on SuSE Linux (if you
noticed, I'm still looking).  Have you ever considered making the
Eiffel GPL installed on one/several linux distros available in that
format?  Anything to help sipping from this firehose and spread the
zen.

> looking).  Have you ever considered making the Eiffel GPL
> installed on one/several linux distros available in that
> format?  Anything to help sipping from this firehose and spread the
> zen.

Indeed, that something we always wanted to do. We are providing a way to
repackage EiffelStudio in a more unix manner (i.e. installed in /usr/bin,
/usr/lib and /usr/share). If you want to help, I'll suggest subscribing to the
es-devel mailing list at
http://origo.ethz.ch/cgi-bin/mailman/listinfo/es-devel).

Thanks,
Manu

Is there a recent release of Eiffel for OS X (10.4) ?
I have found an older one but it complains about ccompile being missing when I
try to compile anything. I worked on it a while, to no avail.
At this point I just want to test drive, not rebuild the engine :-)

Thanks!

Regards,
   JJ

---
Help everyone. If you can't do that, then at least be nice.

Today I used eStudio 5.7 to generate documentation for a little 12
class system.  Very Nice!

When I browsed through the generated documentation, I noticed the
cluster page displays some default headings for documentation
(description and license, I think) but it shows those items as empty
strings.

How does one write the documentation to be used by the documentator
for clusters?  Where can I look to find more developers' detail on the
documentor and the options one has? ... Oh, I'll bet I should be
looking in the eStudio help facility.  But, since I have already asked
the question ...

I may have misunderstood, but I think Ed Mathias suggested to provide
vmware image of linux distribution including EiffelStudio GPL installed
to allow people to try Eiffel easily.
In a similar spirit, a live CD including EiffelStudio and a few popular
Eiffel library would be nice.
But a better place to talk about this is es-devel mailing list at
http://origo.ethz.ch/cgi-bin/mailman/listinfo/es-devel .

Jocelyn

On 11/4/2006 22:59 PM, Emmanuel Stapf [ES] wrote:
>> looking).  Have you ever considered making the Eiffel GPL
>> installed on one/several linux distros available in that
>> format?  Anything to help sipping from this firehose and spread the
>> zen.
>>
>
> Indeed, that something we always wanted to do. We are providing a way to
> repackage EiffelStudio in a more unix manner (i.e. installed in /usr/bin,
> /usr/lib and /usr/share). If you want to help, I'll suggest subscribing to the
> es-devel mailing list at
> http://origo.ethz.ch/cgi-bin/mailman/listinfo/es-devel).
>
> Thanks,
> Manu
>
>

> Is there a recent release of Eiffel for OS X (10.4) ?

At the moment, the latest releases for the Mac can be found here:
http://eiffelsoftware.origo.ethz.ch/index.php/EiffelOnMac
with an explanation on how to get things setup.

Regards,
Manu

> How does one write the documentation to be used by the
> documentator for clusters?  Where can I look to find more
> developers' detail on the documentor and the options one has?
> ... Oh, I'll bet I should be looking in the eStudio help
> facility.  But, since I have already asked the question ...

In the project settings, you have a description entry which is extracted when
generating the documentation.

Manu

> The vertical box contains EV_HORIZIONTAL_BOXes, each of which
> contains an EV_LABEL and an EV_PIXMAP.

I could not reproduce this behavior, it seems to work fine for me. Does it fail
for you on a small example as well?

Regards,
Manu

> Then I checked the value in retrieve_exception and it was 23,
> as if something went wrong. How can that be? What am I missing?

`retrieve_exception' is a constant inherited from EXCEPTIONS and its value is
always 23. You should use the constant when you get an exception to see what
kind of exception it is.

> By the way: what are the known limitations of using STORABLE?
> When I should prefer using the classes in the serialization
> cluster or other storage mechanisms?

STORABLE is using the C implementation that has been around for a few years. It
cannot be used to exchange data with the .NET version. The serialization
clusters is solely written in Eiffel using INTERNAL to offer compatibility with
.NET as well; it does not support serialization of expanded objects, and there
is not yet a mechanism to handle mismatch (aka schema evolution).

Hope this helps,
Regards,
Manu

This should be fixed in the next relase.

Manu

> Thomas Beale wrote:
>
> - checkout the subversion URL -
> http://svn.openehr.org/ref_impl_eiffel/TRUNK/ into
> some\where\ref_impl_eiffel
> - go to apps/adl_workbench/app and build from the windows.ace
> (either set the environment variable OPENEHR=some\where or
> just hack the ace file directly)
> - run the app
> - go to the options menu & select the option setting
> - on the dialog box that you see, hit OK
> - repeat the options menu steps above
> It should cause an assertion violation the second time the
> dialog is displayed.
>
> - thomas
>
> > Manu
> >
> >
> >> -----Original Message-----
> >> From: eiffel_software@yahoogroups.com
> >> [mailto:eiffel_software@yahoogroups.com] On Behalf Of Thomas Beale
> >> Sent: Monday, September 25, 2006 6:22 AM
> >> To: eiffel_software@yahoogroups.com
> >> Subject: Re: [eiffel_software] Vision2 Q - assertion violations on
> >> dialog Ok button
> >>
> >>
> >> I thought it was obvious, since the class names are marked in the
> >> comment blocks of the routines, as generated by ES's own tools;-)
> >>
> >> - thomas
> >>
> >> Peter Gummer wrote:
> >>
> >>> Thomas Beale wrote:
> >>>
> >>>> Sorry, Manu, I don't know what you mean by this!
> >>>>
> >>> Maybe Manu doesn't realise that the code you posted is from
> >>>
> >> somewhere
> >>
> >>> in the
> >>> Vision2 library. It took me a while to figure that out from your
> >>> original question. Initially I thought the code you had
> posted was
> >>> from one of your own classes.
> >>>
> >>> - Peter Gummer
> >>>
> >>>
> >>>
> >> --
> >> ______________________________________________________________
> >> _____________________
> >> CTO Ocean Informatics (http://www.OceanInformatics.biz) Research
> >> Fellow, University College London
> >> (http://www.chime.ucl.ac.uk) Chair Architectural Review Board,
> >> openEHR (http://www.openEHR.org)
> >>
> >>
> >>
> >>
> >>
> >> Yahoo! Groups Links
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
>
>
> --
> ______________________________________________________________
> _____________________
> CTO Ocean Informatics (http://www.OceanInformatics.biz)
> Research Fellow, University College London
> (http://www.chime.ucl.ac.uk) Chair Architectural Review
> Board, openEHR (http://www.openEHR.org)
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>

I am wondering what the effect will be if I have a routine in a
single-threaded Vision2 application that takes a long time to process.
(5-10 secs)  Will I get a "program not responding" error? It is not
possible to break it down because I am calling out to a C routine.

Should I use multi-threading.

Finnian Reilly

Hello Finnian!

You really should run that in a separate thread.  The problem is not so much
"Program not responding"; it's the fact that all UI events will be blocked until
the routine ends.  You won't be able to repaint or move while the calculation is
running and your uses will be quite unhappy with that.

     R

==================================================
Roger F. Osmond
----------------------------------------
Amalasoft Corporation
273 Harwood Avenue
Littleton, MA 01460



  -------- Original Message --------
Subject: [eiffel_software] Vision2 and recommended threading model for
long calculations
From: "Finnian" <finnianr@...>
Date: Wed, November 08, 2006 11:20 am
To: eiffel_software@yahoogroups.com






I am wondering what the effect will be if I have a routine in a
single-threaded Vision2 application that takes a long time to process.
(5-10 secs) Will I get a "program not responding" error? It is not
possible to break it down because I am calling out to a C routine.

Should I use multi-threading.

Finnian Reilly

Hello Roger,
thanks a lot for your tip which makes a lot of sense. Now that I am on
the multi-threaded path my next question is how to notify the GUI
thread when the worker thread has finished. I have seen running a
EV_TIMER suggested as one possible solution but ideally I would like
to find a Vision2 equivalent to the WEL technique of calling
post_message. (simulated button click or some such) post_message
relies on the windows message queue so perhaps there is no equivalent.

Any thoughts on how to do this in Vision2?

Finnian


--- In eiffel_software@yahoogroups.com, rfo@... wrote:
>
> Hello Finnian!
>
> You really should run that in a separate thread.  The problem is not
so much "Program not responding"; it's the fact that all UI events
will be blocked until the routine ends.  You won't be able to repaint
or move while the calculation is running and your uses will be quite
unhappy with that.
>
>     R
>
> ==================================================
> Roger F. Osmond
> ----------------------------------------
> Amalasoft Corporation
> 273 Harwood Avenue
> Littleton, MA 01460
>
>
>
>  -------- Original Message --------
> Subject: [eiffel_software] Vision2 and recommended threading model for
> long calculations
> From: "Finnian" <finnianr@...>
> Date: Wed, November 08, 2006 11:20 am
> To: eiffel_software@yahoogroups.com
>
>
>
>
>
>
> I am wondering what the effect will be if I have a routine in a
> single-threaded Vision2 application that takes a long time to process.
> (5-10 secs) Will I get a "program not responding" error? It is not
> possible to break it down because I am calling out to a C routine.
>
> Should I use multi-threading.
>
> Finnian Reilly
>

I do not think I will be able to reproduce it on a small example, because it
used to work.
So the circumstances are likely to be quite subtle.

Also, I don't know at what point it stopped working, because I mostly
develop this application under Linux.

To reproduce it, you will need a recent version of Gobo from CVS, plus the
2.5 beta of ePOSIX.

Then check out gestalt (start at gestalt.sourceforge.net), change to
directory explorer, and compile with the command:

geant compile_debug_ise


>From: "Emmanuel Stapf [ES]" <manus@...>

> > The vertical box contains EV_HORIZIONTAL_BOXes, each of which
> > contains an EV_LABEL and an EV_PIXMAP.
>
>I could not reproduce this behavior, it seems to work fine for me. Does it
>fail
>for you on a small example as well?

_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters

Hello Finnian!

In the olden days (w/ Eiffel 2, for example), I had my own libraries that
included main loops and shared services that checked for windowing events as
well as activity on sockets.  That changed (thankfully) when Vision (and later
Vision2) came along.  Now, EV_APPLICATION has a main loop (message_loop) that
does at least the windowing message part.  For pure Windows apps, I used to have
a suite of *_IMP classes of my own to handle such things as painting widgets
directly from subthreads (e.g. for a high-resolution frame counter for an on-air
broadcast system).  There is no such portable option in Vision (at least at the
moment - Hint to Manu).

If you think about it a little, there should be a relatively easy way to exploit
the main message loop.  You need not post a message (especially the nearly
useless WM_TIMER message), but you do have to have a look to see if the worker
thread is done from time to time.  The crudest way might be to catch a signal
when the child exits - this is a very bad idea because signals in Eiffel are
treated as exceptions.  You really don't want to go there.

A cleaner, more rational, and more EiffelVision-like way is to take advantage of
the breath that message_loop takes each time through.

EV_APPLICATION provides a means by which to add actions (agents) to call each
time the message loop is idle, or just once if that's appropriate.  These
routines are called 'add_idle_action' and 'do_once_on_idle'.

You create a PROCEDURE to check the status of the worker thread(s), and give
that agent to add_idle_action.  Each time through message_loop when there are no
pending messages, your procedure will be call.  That way, when the thread is
complete, you can either flag it as such for subsequent processing, or simply
capture the value at that time.

I hope this helps you.

      R

==================================================
Roger F. Osmond
----------------------------------------
Amalasoft Corporation
273 Harwood Avenue
Littleton, MA 01460



  -------- Original Message --------
Subject: [eiffel_software] Re: Vision2 and recommended threading model
for long calculations
From: "Finnian" <finnianr@...>
Date: Wed, November 08, 2006 5:10 pm
To: eiffel_software@yahoogroups.com






Hello Roger,
thanks a lot for your tip which makes a lot of sense. Now that I am on
the multi-threaded path my next question is how to notify the GUI
thread when the worker thread has finished. I have seen running a
EV_TIMER suggested as one possible solution but ideally I would like
to find a Vision2 equivalent to the WEL technique of calling
post_message. (simulated button click or some such) post_message
relies on the windows message queue so perhaps there is no equivalent.

Any thoughts on how to do this in Vision2?

Finnian

--- In eiffel_software@yahoogroups.com, rfo@... wrote:
>
> Hello Finnian!
>
> You really should run that in a separate thread. The problem is not
so much "Program not responding"; it's the fact that all UI events
will be blocked until the routine ends. You won't be able to repaint
or move while the calculation is running and your uses will be quite
unhappy with that.
>
> R
>
> ==================================================
> Roger F. Osmond
> ----------------------------------------
> Amalasoft Corporation
> 273 Harwood Avenue
> Littleton, MA 01460
>
>
>
> -------- Original Message --------
> Subject: [eiffel_software] Vision2 and recommended threading model for
> long calculations
> From: "Finnian" <finnianr@...>
> Date: Wed, November 08, 2006 11:20 am
> To: eiffel_software@yahoogroups.com
>
>
>
>
>
>
> I am wondering what the effect will be if I have a routine in a
> single-threaded Vision2 application that takes a long time to process.
> (5-10 secs) Will I get a "program not responding" error? It is not
> possible to break it down because I am calling out to a C routine.
>
> Should I use multi-threading.
>
> Finnian Reilly
>

I want to use MySQL in an eiffel system.  I have found a few reported
facilities for MySQL in Eiffel:
1) By M. Ravits, on sourceforge.  I have pulled down the sources for
this wrapper but it does not compile on ISE5.7.  I might be able to
make this work, but wouldn't I rather not?
2) By elj, also on sourceforge, although I am unable to get a
connection to the source through CVS, or the example ("connection
refused")
3) Then there is the Eiffel Wrapper Generator ... I suppose I could
try to wrap the C language MySQL driver for use by eiffel ... But
wouldn't I rather not?

4) Some other, more current and fully functional facility ??

Any help, please?

ECLI.



>
>4) Some other, more current and fully functional facility ??

_________________________________________________________________
Windows Live™ Messenger has arrived. Click here to download it for free!
http://imagine-msn.com/messenger/launch80/?locale=en-gb

>EV_APPLICATION provides a means by which to add actions (agents) to call
>each time the message loop is idle, or just once if that's appropriate.
>These routines are called 'add_idle_action' and 'do_once_on_idle'.
>

add_idle_action appears to noly be in EV_APPLICATION_I (at least in 5.6 -
maybe this has changed in 5.7).

There is also EV_TIMEOUT.

_________________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters

Hello Robert,

There are other solutions :
* EiffelStore + ODBC Handle.
* ECLI - safe.sourceforge.net (Eiffel library for relational database
access through ODBC).
* EPOM(Eiffel Persistent Object Management) + ECLI  -
safe.sourceforge.net

In both cases check the following :
- a MySQL ODBC driver is installed on your computer
- an ODBC datasource has been defined, as a link to your actual MySQL
database.

Best regards,

Paul G. Crismer



________________________________

	 From: eiffel_software@yahoogroups.com
[mailto:eiffel_software@yahoogroups.com] On Behalf Of robert.marcum
	 Sent: jeudi 9 novembre 2006 14:50
	 To: eiffel_software@yahoogroups.com
	 Subject: [eiffel_software] Which Eiffel MySQL facility?



	 I want to use MySQL in an eiffel system. I have found a few
reported
	 facilities for MySQL in Eiffel:
	 1) By M. Ravits, on sourceforge. I have pulled down the sources
for
	 this wrapper but it does not compile on ISE5.7. I might be able
to
	 make this work, but wouldn't I rather not?
	 2) By elj, also on sourceforge, although I am unable to get a
	 connection to the source through CVS, or the example
("connection
	 refused")
	 3) Then there is the Eiffel Wrapper Generator ... I suppose I
could
	 try to wrap the C language MySQL driver for use by eiffel ...
But
	 wouldn't I rather not?

	 4) Some other, more current and fully functional facility ??

	 Any help, please?







***** Disclaimer *****
http://www.groupes.be/1_mail-disclaimer.htm

[Non-text portions of this message have been removed]

Hello Colin (et al)!

You are correct.  In 5.6, you could find 'add_idle_action' in EV_APPLICATION_I,
so you would have to call it 'implementation.add_idle_action'.  In 5.7, the call
exists in EV_APPLICATION itself, so the call through implementation is no longer
necessary.

EV_TIMEOUT is handy enough, but at least on Windows, it's incipid. It seems to
use WM_TIMER events.  This is the norm for "low resolution" timing on Windows
and has some interesting behaviors.  WM_TIMER events have the lowest priority of
any events in Windows.  Therefore, there is really no certainty at all as to
when the actions associated with the these events will occur.  If, for example,
there are any mouse events queued WHILE the WM_TIMER event is in the message
queue, the WM_TIMER event will be moved to the back of the queue.
For cases where a little more predictability is required, it is necessary to
paint directly to a control from a subthread.  This is quite platform-dependent
as you might imagine, and I have resorted to direct C calls in such cases.

What I challenge the Eiffel community to do is to solve this in a portable,
Eiffel-like way.  There must be equivalents in gtk, so the *_IMP mechanism we
all know and love could be used to provide this sort of thing.  Sure, it would
have to be multithreaded (and threading is still not as solid as some of us
would like), but that's pretty much the norm for non-trivial GUIs anyway.

My current challenge is to be able to animate figures in an EV_MODEL_WORLD in an
efficient (and smooth) way.  Working with EV_TIMEOUT is just not very
satisfactory.  20 frames per second means a timeout of 50 msecs
(nondeterministic timing notwhithstanding) and that coupled with the
projector.project calls just hammers the CPU.
A multithreaded alternative strikes me as a major Pandora's box.  How, for
example are the world and projector shared (and locked) amongst the threads? 
Even if all of this messiness is solved, would it even help?  My guess is that
the mechanism is so inefficient that it's not worth the trouble.  Can somebody
out there prove me wrong?  I'd like that a lot.

Thanks,
      R

==================================================
Roger F. Osmond
----------------------------------------
Amalasoft Corporation
273 Harwood Avenue
Littleton, MA 01460



  -------- Original Message --------
Subject: RE: [eiffel_software] Re: Vision2 and recommended threading
model for long calcu
From: "Colin Adams" <colinpauladams@...>
Date: Thu, November 09, 2006 8:57 am
To: eiffel_software@yahoogroups.com






>EV_APPLICATION provides a means by which to add actions (agents) to call
>each time the message loop is idle, or just once if that's appropriate.
>These routines are called 'add_idle_action' and 'do_once_on_idle'.
>

add_idle_action appears to noly be in EV_APPLICATION_I (at least in 5.6 -
maybe this has changed in 5.7).

There is also EV_TIMEOUT.

__________________________________________________________
Be the first to hear what's new at MSN - sign up to our free newsletters!
http://www.msn.co.uk/newsletters