JWS/JOGL test..

[Andrew Thompson]

I am attempting to launch previews of JOGL[1] based
Saverbeans[2] screensavers.

A test on one of them has returned mixed results and I
would appreciate further test results.

Note that it takes around 6 Megs, Java 1.5, and accepting
signed code (by this outfit called 'Sun') to get this on-screen.
<http://www.javasaver.com/testjs/pv/pv.html#javasaver>

Click the 'planetview.jnlp' link and report as to which
of the two images (shown at the top of that page) more
closely describes what appears on-screen. I refer to the
second as 'white blob'.

[1] <https://jogl.dev.java.net/>
[2] <https://screensavers.dev.java.net/>

 

[Roedy Green]

<http://www.javasaver.com/testjs/pv/pv.html#javasaver>

This displayed a screen with two images on it, then quickly yanked
them away before I could examine them. On the following text page I
clicked the planetview.jnlp link.

It then loaded over a meg of stuff from two different websites giving
me an irritating progress message that gave no indication of how long
this was going to go on.

Upon Java's recommendation, I accepted the Sun cert, and off it went
and displayed a black and gold globe smoothly spinning much faster
than real time.

I am using Win2K, and Java 1.5_05.

 

[Andrew Thompson]

This displayed a screen with two images on it, then quickly yanked
them away before I could examine them.

....scroll up. (note the 'anchor' on the end of the URL?

...On the following ..

same actually.

..text page I
clicked the planetview.jnlp link.

It then loaded over a meg of stuff from two different websites giving
me an irritating progress message that gave no indication of how long
this was going to go on.

I specify the size of the jar files in the JNLP I created,
in case the server does not return the file size, that is
what the client side Webstart uses for the 'indicator bars'
of both the saverbeans-api.jar (tiny ~40Kb), and the
planetview.jar.jar (around 2 meg).

I have observed the progress for the larger file, which at
first reports 'nnnK of 2.1 Meg' downloaded, then it shows
the prgress bar, then, once an approximate *speed* is
established, it gives a 'time till completion' estimate.

The 40K jar is too small to invoke most of that behaviour
(on my set-up).

Unfortunately, all the JOGL jars are called from a *different*
JNLP[1] at the JOGL project, and since tit (apparently) does not
list the jar sizes, and the JDIC site (apparently) does not
return the file sizes, the progress dialogs will not appear.

That is a pity, and I am tempted to approach them directly
to get that changed. Wish me luck.

[1] I forogt to mention, the other day when we were discussing
the <applet-desc> element in the JNLP syntax, I also discovered
two much more useful ones - <component-desc> and <installer-desc>.
JOGL is marked as a component, so I can reference it from my
site in a JNLP as an 'extension'. It saves me building, uploading
the (big) JOGL files, but also the bandwidth of serving them.
Unfortunately, it does also reduce the control I have..

Upon Java's recommendation, I accepted the Sun cert, and off it went
and displayed a black and gold globe

Must be Sol(?).

Are you positive it is 'gold' rather than 'white'?

Here is a (cough)direct(cough) link to those screenshots.

..smoothly spinning much faster
than real time.

Aha! That info. is probably useful to the author, I'll
pass it on.

I am using Win2K, and Java 1.5_05.

Thanks for the test results Roedy, as well as the detailed
description of your experiences. With that in mind, I would
make slight changes to future posts (a little more info.,
fewer HTML 'anchors'.

 

[Roedy Green]

I specify the size of the jar files in the JNLP I created,
in case the server does not return the file size, that is
what the client side Webstart uses for the 'indicator bars'
of both the saverbeans-api.jar (tiny ~40Kb), and the
planetview.jar.jar (around 2 meg).

I did not get any indicator bars, just an increasing numeric without
an indication of where it was headed. I know this is not your code,
but it is something not right.

 

[Roedy Green]

I have observed the progress for the larger file, which at
first reports 'nnnK of 2.1 Meg' downloaded, then it shows
the prgress bar, then, once an approximate *speed* is
established, it gives a 'time till completion' estimate.

I saw nothing like that. Perhaps this is a browser problem. I was
using Opera 8.5. I will try again with some other browsers.

How big a file can http headers report?

 

[Roedy Green]

I have observed the progress for the larger file, which at
first reports 'nnnK of 2.1 Meg' downloaded, then it shows
the prgress bar, then, once an approximate *speed* is
established, it gives a 'time till completion' estimate.

tried with IE and Win2k, after uninstalling JAWS planet app to force
reload.

Same progress behaviour
loading jarname... from website...
1.7M loaded

no progress bars, no ultimate size.

 

[Roedy Green]

I have observed the progress for the larger file, which at
first reports 'nnnK of 2.1 Meg' downloaded, then it shows
the prgress bar, then, once an approximate *speed* is
established, it gives a 'time till completion' estimate.

same again with Firefox and Win 2k. The only difference was there is
one additional message when the JNLP file is finished loading in the
bottom right of the screen.

 

[Andrew Thompson]

I did not get any indicator bars, ..

...please note that
- There is only one file of any size that has the proper
file size indicator.
- I might set up an independent demonstration using *only* files
that have a size specified in the JNLP.
- This test is *not* about JWS progress bars!

 

[Chris Uppal]

Click the 'planetview.jnlp' link and report as to which
of the two images (shown at the top of that page) more
closely describes what appears on-screen. I refer to the
second as 'white blob'.

For me both jnlp links produce the same result: a rotating world rather than a
white blob.

That's with Java 1.5.0, but I'm afraid I have little idea about the version of
OGL on this machine, and that -- I assume -- is the more relevant info.
Anyway, it's the one that comes with WinXP, the opengl32.dll is version
5.1.2600.1106, if that helps ;-)

BTW, it also uses 100% CPU and is apparently running at high enough priority
that other apps become unresponsive, and it's hard to kill itself :-(

-- chris

 

[Andrew Thompson]

For me both jnlp links produce the same result: a rotating world rather than a
white blob.

Thank you! Two 'works', and two 'no work'*.

* I am guessing Roedy's 'golden' was actually a 'no work'.

That's with Java 1.5.0, but I'm afraid I have little idea about the version of
OGL on this machine, and that -- I assume -- is the more relevant info.

No. The JOGL stuff used for the demo *should* be
coming from jogl.dev.java.net.

Anyway, it's the one that comes with WinXP, the opengl32.dll is version
5.1.2600.1106, if that helps ;-)

....hmm. I'm showing version 5.1.2600.0 for that DLL.
(I don't think it is relevant, but it is hard to say
for sure before the problem is resolved.)

BTW, it also uses 100% CPU and is apparently running at high enough priority
that other apps become unresponsive, and it's hard to kill itself :-(

Sorry, they are intended as Screensavers, and depend on an
underlying API that calls a repaint at a rate of 60Hz (AFAIR).
It makes WinAmp (which I've heard play flawlessly during
system lock-ups, and otherwise extremely heavy useage)
'miss a beat'.

This is all *very* experimental.

See how you go with these two savers. The first is not JOGL
and only requires a download of <100Kb.
<http://www.javasaver.com/testjs/xsl/04/jws/install.html>
Note that the Webstarted 'Applet' links below those, suck.
( JWS and applets are a bad combo., I have decided.

 

[Chris Uppal]

* I am guessing Roedy's 'golden' was actually a 'no work'.

Well, the globe does look fairly golden -- all those deserts -- so it may be
that Roedy was reporting a 'work'. (Admittedly there's more sea than desert,
but the overall impression is rather golden -- at least to a land animal, I
suppose a fish would say it was almost all blue).

No. The JOGL stuff used for the demo *should* be
coming from jogl.dev.java.net.

Isn't that built on the underlying OGL implementation in the OS (if there is
one) ? Certainly the downloaded DLL seems to use the system opengl32.dll
(according to DependencyWalker).

...hmm. I'm showing version 5.1.2600.0 for that DLL.
(I don't think it is relevant, but it is hard to say
for sure before the problem is resolved.)

I just tried on a Win2K box, with a freshly installed JDK 1.5.0_5. That box
has version 5.0.2195.6611 of the opengl32 DLL. Both jnlps showed a rotating
globe. I think that machine must be better at using hardware acceleration in

See how you go with these two savers. The first is not JOGL
and only requires a download of <100Kb.
<http://www.javasaver.com/testjs/xsl/04/jws/install.html>

The line demo runs at more-or-less zero CPU (even if I maximise the window !!).
The picture cube is just as bad as the rotating world. (And, for some reason
downloaded JOGL yet again [*]).

-- chris

([*] making it five downloads of JOGL so far this morning -- if anyone's
monitoring my net habits they must think I'm some kind of pervert...)

 

[Andrew Thompson]

Andrew Thompson wrote:

Isn't that built on the underlying OGL implementation in the OS (if there is
one) ? Certainly the downloaded DLL seems to use the system opengl32.dll

'the' system version? I understand that JOGL distributes some
'natives' to get things working.

(according to DependencyWalker).

You now know better than I do, so.. apparently. ;-)

I just tried on a Win2K box, with a freshly installed JDK 1.5.0_5. That box
has version 5.0.2195.6611 of the opengl32 DLL. Both jnlps showed a rotating
globe. I think that machine must be better at using hardware acceleration in
the OGL implementation since CPU was reasonable rather than <ouch!>.
Interesting..


The line demo runs at more-or-less zero CPU (even if I maximise the window !!).
The picture cube is just as bad as the rotating world.

OK - non JOGL low CPU, JOGL high.

..(And, for some reason
downloaded JOGL yet again [*]).

I will have to look into that. The JOGL API is marked

([*] making it five downloads of JOGL so far this morning -- if anyone's
monitoring my net habits they must think I'm some kind of pervert...)

Or a bandwidth HOG(L). ;-)

Thanks for all the downloads you've suffered, and thanks
for the feedback and results.

 

[Scott Ellsworth]

Click the 'planetview.jnlp' link and report as to which
of the two images (shown at the top of that page) more
closely describes what appears on-screen. I refer to the
second as 'white blob'.

Rotating world rather than white blob.

MacOS X 10.4, 1.5.0_05

work@boggle:build$ java -version
java version "1.5.0_05"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_05-64)
Java HotSpot(TM) Client VM (build 1.5.0_05-43, mixed mode, sharing)

 

[Roedy Green]

* I am guessing Roedy's 'golden' was actually a 'no work'.

I was a gold and black representation of planet earth. It looks
plausible as a desired result. Perhaps you could post an JPG of what
you want it to look like.

 

[Roedy Green]

* I am guessing Roedy's 'golden' was actually a 'no work'.

I have emailed you a screen capture so you can judge for yourself.

 

[Andrew Thompson]

I have emailed you a screen capture so you can judge for yourself.

Thank you Roedy! I appreciate the effort you put into
sending that screenshot to me[1], and it certainly confirms
that you are seeing 'rotating planet', rather than the
'white blob' I am seeing.

[1] I could bounce the screenshot I am seeing back to you, but..
a) There is no further doubt in my mind that you are seeing
the 'correct thing'.
b) It is truly, truly boring. It is a big, blob of undifferentiated
white with a few stars around it. The only thing that gives a sense
of rotation are the rotating BG of stars (very subtle, hard to notice)
and the outline of the 'blob of white' changes slightly.

 

[Andrew Thompson]

Rotating world rather than white blob.

MacOS X 10.4, 1.5.0_05

Heh.. Way cool! The SaverBeans savers themselves are not
yet available for Mac's, but I have been looking for 'leverage'
to convince the author to port the underlying API to it.

This may help, as he defected (from Linux) to Mac. some
months ago.

 

[Andrew Thompson]

I am attempting to launch previews of JOGL[1] based
Saverbeans[2] screensavers.

Halt! The author of the saver gave me a development
version where it was easier to configure the 'image detail'
provided in the screen rendering.

This PC was able to render the planet on 'low resolutions',
but the image went to 'white blob' at higher resolutions.

This gives us a new direction to research, and I suspect
gives the author the reason why it is only failing on one
(crappy) PC.

Thanks to all respondents (and anybody who took the interest
to read!).

 

[Roedy Green]

This gives us a new direction to research, and I suspect
gives the author the reason why it is only failing on one
(crappy) PC.

One thing to check is the version of DirectDraw installed.

 

[Andrew Thompson]

I am attempting to launch previews of JOGL[1] based
Saverbeans[2] screensavers.

A test on one of them has returned mixed results and I
would appreciate further test results.

Note that it takes around 6 Megs, Java 1.5, and accepting
signed code (by this outfit called 'Sun') to get this on-screen.
<http://www.javasaver.com/testjs/pv/pv.html#javasaver>

Problem solved.
- My own (crappy) PC* was the only PC unable to render the
saver at the default options
- Tests with a lower resolution, 'imagedetail 2', work on this PC.
- Those who can see it at it's default resolution might be
interested to see it at 'imagedetail 2' for comparison,
so I have added links to JNLP's that add it to the options
(search for 'imagedetail 2' in the page).
- The JOGL jar's definitely seem to be cached here, now,
I no longer have to wait more than a few moments to see
the previews.

* AMD Athlon XP1800+ with 1 Gig memory and ..NVIDIA
GeForce2 MX 100/200 video card (shrugs vaguely).