Dolphin 5.1 Release Candidate

Folks,

We have a Dolphin 5.1 Release Candidate for you to try out if you wish. The
5.1 upgrade will be free for all existing users of Dolphin Smalltalk XP and
your current serial numbers should be used to unlock it. We are hoping that
this will be the final release but would like to elicit some user feedback
first (just to make sure there are no genuine show-stoppers in the software
or install procedure).

You can download the files using these (temporary) links:

http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DPROSetup51.exe
http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DSESetup51.exe
http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DVESetup51.exe

The release notes can be found at:

http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/ReleaseNotes51.htm

Please take a look at these to find out what is supposed to be
fixed/improved in this version. The code is effectively frozen so please
don't complain if the particular bug or feature you were looking forward to
is not in this release. In terms of feedback, what we want to know is:

(a) Have we fixed what we think we have?
(b) Have we broken anything that wasn't broken in the previous version?
(c) Does the install procedure work correctly? In particular, you should be
able to install and run 5.1 alongside all other versions of Dolphin without
problems.

Please don't think that you have to upgrade to this Release Candidate right
away. If you would prefer to wait until the final 5.1 is out please feel
free to do so.

Best Regards,

Andy Bower
Dolphin Support
http://www.object-arts.com
---
Are you trying too hard?
http://www.object-arts.com/Relax.htm
---

Hi Andy,

> Please take a look at these to find out what is supposed to be
> fixed/improved in this version. The code is effectively frozen so please
> don't complain if the particular bug or feature you were looking forward
to
> is not in this release. In terms of feedback, what we want to know is:
>
> (a) Have we fixed what we think we have?
> (b) Have we broken anything that wasn't broken in the previous version?
> (c) Does the install procedure work correctly? In particular, you should
be
> able to install and run 5.1 alongside all other versions of Dolphin
without
> problems.

Installed the pro version without a problem and loaded and ran several of my
apps and tests without a problem.

Cheers,

Ted
www.tedbracht.co.uk

Andy,

Can you explain a little about 1228?  If you are adding more of the
behavior, why is it a good thing?  I'm just picturing release day for one of
my apps, running up and down stairs a few times, and maybe doing it once
more because I didn't set the clocks quite right.

Please relay congratulations and thanks to Blair and Timo.

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Andy,

The installation seems to work correctly.

It's not really something that you want to hear about 5.1, but<g>, in
looking at whether there might be a performance hit from the changes to
Fraction #= and #hash, I noticed #asFloatD, #asFloatE and #asFloatQ in
Float.  They appear to be unreferenced; they are also present in my 5.0
image.

A few release notes point to things that might require some changes on my
part, but nothing that looks unpleasant, and many will be quite helpful.

I was a little surprised to see a fix for a cosmetic Windows XP bug (the
list view thing).  OTOH, I'm not at all surprised to see you (yet again)
providing better support than Microsoft.  However, once something like this
gets in the image, it could be some time before anyone notices it's no
longer needed.  Is there an easy way to do the OS version test once?
Perhaps a block that gets set on startup?

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Bill,

> Can you explain a little about 1228?  If you are adding more of the
> behavior, why is it a good thing?  I'm just picturing release day for one
of
> my apps, running up and down stairs a few times, and maybe doing it once
> more because I didn't set the clocks quite right.

I think you may be misunderstanding the #1228 fix (either that, or I haven't
the faintest idea what you mean here). The description of #1228 in the
Release Notes is a description of the problem that has been fixed. Perhaps
you were reading it as a new feature that has been added?

> Please relay congratulations and thanks to Blair and Timo.

Since they both read this newsgoup, I think you just have.

Best Regards,

Andy Bower
Dolphin Support
http://www.object-arts.com
---
Are you trying too hard?
http://www.object-arts.com/Relax.htm
---

Bill

You wrote in message news:b6nkks$v0b$[hidden email]...
> Andy,
>
> The installation seems to work correctly.
>
> It's not really something that you want to hear about 5.1, but<g>, in
> looking at whether there might be a performance hit from the changes to
> Fraction #= and #hash, I noticed #asFloatD, #asFloatE and #asFloatQ in
> Float.  They appear to be unreferenced; they are also present in my 5.0
> image.

These are present for ANSI compatibility (note that they are part of the
ANSI <number> protocol).

>
> A few release notes point to things that might require some changes on my
> part, but nothing that looks unpleasant, and many will be quite helpful.
>
> I was a little surprised to see a fix for a cosmetic Windows XP bug (the
> list view thing).  OTOH, I'm not at all surprised to see you (yet again)
> providing better support than Microsoft.  However, once something like
this
> gets in the image, it could be some time before anyone notices it's no
> longer needed.  Is there an easy way to do the OS version test once?
> Perhaps a block that gets set on startup?

Probably, by using a bit of polymorphism, but since this test is only
performed once at the end of each vertical scroll operation, and seems to
run in less than a microsecond on this machine (the OS name is cached once
determined the first time), the current approach seems adequate.

Regards

Blair

Andy,

> > Can you explain a little about 1228?  If you are adding more of the
> > behavior, why is it a good thing?  I'm just picturing release day for
one
> of
> > my apps, running up and down stairs a few times, and maybe doing it once
> > more because I didn't set the clocks quite right.
>
> I think you may be misunderstanding the #1228 fix (either that, or I
haven't
> the faintest idea what you mean here). The description of #1228 in the
> Release Notes is a description of the problem that has been fixed. Perhaps
> you were reading it as a new feature that has been added?

That was it - it read more like a deliberate change than a fix.  Sounds
great!

Have a good one,

Bill

--
Wilhelm K. Schwab, Ph.D.
[hidden email]

Andy,

I've just had a quick play with D5.1rc1. It seemed pretty hitchless
(though I'll prod it a little harder later on).

One new bug I noticed (which you may well not think worth fixing at this time).
Debugging into:
    [:cc | cc value: true] callCC.
gives a walkback ('Debugger does not understand #return:toFrame:').

Also noticed an oddity with workspaces, and I wonder if anyone else is seeing
the same thing. Closing a workspace with changed text gives the usual prompt to
save the changes, but in 5.1 it seems to take noticeably longer to put up the
prompt than it did in 5.0. It's variable but feels like about a second most
often (this is a fairly slow machine). Closing other D5.1 windows happens as
fast as I expect. 5.0 is not noticeably slow about this. I took a very quick
look at the code but didn't see anything strange. I also checked for the
"extra" process that Ian mentioned and there doesn't seem to be one.

-- chris

Chris,

> Also noticed an oddity with workspaces, and I wonder if anyone else
> is seeing the same thing.

Neither of my machines (one using XP, the other 2000) seems at all slow
in opening up the warning dialog.

--
Ian

Andy Bower wrote:

> I think you may be misunderstanding the #1228 fix (either that, or I haven't
> the faintest idea what you mean here). The description of #1228 in the
> Release Notes is a description of the problem that has been fixed. Perhaps
> you were reading it as a new feature that has been added?

In defense of (the other) Bill, #1228 was identified as an "Enhancement". In
general,  your listing of enhancements cite the new, positive thing, that has
been added. Whereas "Incorrect Functionality" or "Crash" items cite the old,
negative thing, that has (by implication) been fixed. Since this particular one
was documented like a bug, but listed as an enhancement, it's easy to see how
one might get confused by what it was saying.

I too have sometimes been confused by one of the descriptions, when it's not
entirely clear whether it's talking about the way things were, or the way that
it behaves now. Perhaps in the future it would help to be just a bit more
explicit and/or consistent in the descriptions?

regards,
-Bill

-------------------------------------------
Bill Dargel            [hidden email]
Shoshana Technologies
100 West Joy Road, Ann Arbor, MI 48105  USA

Ian Bartholomew wrote:
>
> > Also noticed an oddity with workspaces, and I wonder if anyone else
> > is seeing the same thing.
>
> Neither of my machines (one using XP, the other 2000) seems at all
> slow in opening up the warning dialog.

Hmm, odd.

I've just tried instrumenting the time taken between the call to
View>>wmClose:wParam:lParam: and the resulting invocation of
UserLibrary>>overlappedMsgBoxIndirect: from MessageBox>>basicOpen (just after
the call to #forkMainIfMain).  The time is essentially identical on both D5.0
and D5.1 (at less than a millisecond).  Yet there's a very perceptable delay of
1/2 to 1 second on D5.1 before the prompt actually appears.

I don't know of any way to link my "end timer" to the time when the message box
is actually displayed, unfortunately.  Do you (or anyone) ?

As it stands, it looks as if there might be something a bit iffy about
overlapped calls in the new VM. Or #forkMainIfMain.  Or something...

I'm running W2K sp3 on a P3 650Mhz machine.  A similarly (but not identically)
configured W2K box doesn't show the problem, but that's *much* faster machine
(and the TaskManager histogram hints that D5.1 is burning CPU in the interval,
so machine speed may be relevant).

    -- chris

Chris,

> I don't know of any way to link my "end timer" to the time when the
> message box is actually displayed, unfortunately.  Do you (or anyone)
> ?

Not offhand, it's not something I have felt the need to do before.  The
comment for MessageBox>>systemModal might give some hope though?

> I'm running W2K sp3 on a P3 650Mhz machine.  A similarly (but not
> identically) configured W2K box doesn't show the problem, but that's
> *much* faster machine (and the TaskManager histogram hints that D5.1
> is burning CPU in the interval, so machine speed may be relevant).

My Win2K (sp3) box is a P3 @ 750MHz - pretty close to your problem
machine.  As far as I can judge the instant I release the mouse button
to close the workspace the message box pops up - it's that speedy.  I
get the same response using Alt-F4 or the File menu.

--
Ian

Folks,

A new Dolphin 5.1 Release Candidate is now available. This one fixes the
"extra process" problem that Ian Bartholomew found with the previous
version. You MUST uninstall the previous 5.1RC before trying to install the
new one. Once again, we hope this will become the final 5.1 release a few
days from now unless any showstoppers are found.

You can download the installers using these (temporary) links:

http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DPROSetup51.exe
http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DSESetup51.exe
http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/DVESetup51.exe

The release notes can be found at:

http://www.object.arts.dial.pipex.com/Lib/Downloads/5.1RC/ReleaseNotes51.htm

Best Regards,

Andy Bower
Dolphin Support
http://www.object-arts.com
---
Are you trying too hard?
http://www.object-arts.com/Relax.htm
---

Andy,

> A new Dolphin 5.1 Release Candidate is now available.

If you deploy the HelloWorld sample (for example) with the latest RC
then when you run it you get a warning dialog "PC1Cipher class does not
understand #churnRandPool".

--
Ian

Ian,

> > A new Dolphin 5.1 Release Candidate is now available.
>
> If you deploy the HelloWorld sample (for example) with the latest RC
> then when you run it you get a warning dialog "PC1Cipher class does not
> understand #churnRandPool".

Sigh...

Thanks Ian. I've pulled the download from the site pending further
investigation.

Best Regards,

Andy Bower
Dolphin Support
http://www.object-arts.com
---
Are you trying too hard?
http://www.object-arts.com/Relax.htm
---

Ian,

> > I don't know of any way to link my "end timer" to the time when the
> > message box is actually displayed, unfortunately.  Do you (or
> > anyone) ?
>
> Not offhand, it's not something I have felt the need to do before.
> The comment for MessageBox>>systemModal might give some hope though?

That was a good thought.  It doesn't pin down what't happening though.
Instrumenting that shows that it takes -- normally -- about 70msec from the
time a MessageBox is invoked to the time when the parent view looses
activation.  The delay occurs *after* that time.  In fact I can see that the
delay happens after the view looses its "active window" colour coding.  One odd
thing is that about 1 time in 10 that interval is only about 20msec, and in
those cases the message box appears with it's normal (D5.0) speed.

> My Win2K (sp3) box is a P3 @ 750MHz - pretty close to your problem
> machine.  As far as I can judge the instant I release the mouse button
> to close the workspace the message box pops up - it's that speedy.  I
> get the same response using Alt-F4 or the File menu.

That's how it works for me on D5.0.  Not 5.1, though.  It seems that all non
task/system-modal MessageBoxes show the same problem.  Just doing:
    MessageBox notify: 'Hi'.
shows the same syndrome.

I've been looking to see what might be causing this, but have hit a blank.
Dolphin is definitely consuming CPU during the anomalous interval (I can see
the cumulative CPU time in Task Manager going up much faster than it does with
D5.0), but I have no idea what it's doing with the time.  It's not starting up
OS threads (for the "overlapped call" thread pool) -- I can see that in Task
Manager too, and the slowdown doesn't correlate with that.  It's not doing
anything odd with the filesystem or the registry -- the tools from
www.sysinternals.com show that.  So what is it doing ??  Beats me...

    -- chris

Chris Uppal wrote:

Here's some information on a problem that I ran into recently that I believe has
direct bearing on the issue that you're seeing.

The problem I had was on 5.0.3, BUT it did include the patch to
MessageBox>>basicOpen that boosts the priority of the process doing the non task
modal overlapped call. And since that change was my suggestion, I guess I deserve
some of the credit :-( for the problem you're reporting.

My recent problem created a total lockup of Dolphin with 100% cpu consumption. I
had been opening a warning MessageBox, and then after confirmation, the code would
go on to close a shell which would cause a second MessageBox to pop up to ask about
unsaved changes. The second MessageBox wouldn't appear at all, all the while
burning the cpu. Sounds like just an extended version of the delay that you're
seeing?

A control-break would regain control of the UI. The interrupted main (5) process is
stuck in ProcessorScheduler>>callback:return: waiting for "a more recent callback
which has still not returned". Based on the method comment, I guess I must have
inadvertently set up a deadlock somehow? And the 100% processor spin is probably
due to the idle loop spin problem?

I had tried a simple test of two MessageBoxes in succession, but that didn't show a
problem. It must need something more complex going on that causes other callbacks
to occur at the "wrong" time, or something. I had since moved on to replacing the
first message box with signaling an error (which was actually better anyway) and
the problem had gone away.

Chris - Try using the old MessageBox>>basicOpen without the priority boost, and see
if that makes your problem go away. (I just found out that it would fix the problem
I had seen).

Blair - Hopefully this gives you enough information to figure out what's going on?
If not, I can try and see if I can pull out the problem I was having into a test.
Or at least provide more details from the situation in sito. I still think that the
priority boost would be a good thing, provided that these further ramifications can
be addressed.

regards,
-Bill

-------------------------------------------
Bill Dargel            [hidden email]
Shoshana Technologies
100 West Joy Road, Ann Arbor, MI 48105  USA

Bill Dargel wrote:

> Chris - Try using the old MessageBox>>basicOpen without the priority
> boost, and see if that makes your problem go away. (I just found out
> that it would fix the problem I had seen).

Sorry Bill, I'm confused here.  The syndome I'm seeing certainly sounds as if
it could be a result (in some way) of the kind of changes you are talking
about, but I don't think there was a change to MessageBox>>basicOpen between
D5.03 and D5.1.  I can't see one, anyway.

    -- chris

Chris Uppal wrote:

> Sorry Bill, I'm confused here.  The syndome I'm seeing certainly sounds as if
> it could be a result (in some way) of the kind of changes you are talking
> about, but I don't think there was a change to MessageBox>>basicOpen between
> D5.03 and D5.1.  I can't see one, anyway.

Oops, my bad. I haven't actually gotten 5.1 yet, and I guess I must have made an
assumption. [And you know what they say about when you ass-u-me ? ;-/ ]. I
thought I remembered something in the Release Notes, which I had looked through,
but apparently not.

Back in February I had asked (of the MessageBox priority change) "Should it be
included in the base?" to which Blair responded "Yes, I think it should. Its not
ideal, but like I say, a pragmatic solution." I must have translated that
exchange into thinking that it was part of 5.1, especially given the similarity
of the situation you were reporting.

-Bill

-------------------------------------------
Bill Dargel            [hidden email]
Shoshana Technologies
100 West Joy Road, Ann Arbor, MI 48105  USA