Since I've observed this behaviour only on the new CPDN Client on my Win98 machine, I'll post it here as well :

It looks like (once running) my BOINC V4.09 actually can not stop/pause/suspend the running CPDN Client anymore.

When the Machine timeslices to another Project, the CPDN Client is naturally assigned the Status "Paused"

However :
The CPDN Client still runs along, despite being Paused (?!?)

It seems to run effectively on very low Priority compared to the real active Project, resulting in a permanent, but oozingly slow progress in this "Mode".

Since this is the first time I ever observe any Client still running while it actually should be Suspended/Paused, I assume it may contain a Bug that prevents it from stopping/suspending correctly.

I've tried disabling the "Leave Application in Memory" Option for all Projects this machine is running, but (after a manual "Update Project") this did not prevent the CPDN Client from exhibiting this faulty behaviour.

It's funny seeing a "Paused" Project staying active and running along, albeit excessively slow, since the main Project does appear to get most of the CPU cycles assigned. Runtimes of the normal, actively running Project are affected only somewhat (some 25-30% slower than it should run).

-- edit --
Okidok, rechecked all other machines, and the CPDN Clients (Linux and Windows) do properly suspend/pause on all of them.

Only the CPDN Win32 Client 4.04 on the Win98 machine exhibits this problem, so I reckon this is a definitive Bug.

Since the Win98 System is the only one running the new V4.04 Win32 Core, I am not 100.0% sure if only Win98 is affected though.
(I assume during Trickling to the server, the existing V4.03 Win32 Clients do not self-update before finishing a Model, thus I won't see them under the Win2000/WinXP environments I got running as well)

As the Client is this way continually getting CPU time (even while Paused), it might not be timeslicing at all anymore, resulting in permanent, very poor runtime performance due to low efficiency/priority.

-- edit 2 --
The machine did Timeslice back to CPDN, and the other Paused Project (SETI) is behaving normal (not running, no errors observed).
CPDN now gets the full CPU cycles, and predicted total time is naturally dropping every Minute.

I'll wait what happens when CPDN is timesliced (paused) again, and see if it still keeps running in that condition.

-- edit 3 --

Still happens after timeslicing among other Projects, the CPDN Client just keeps running despite being paused :p

___________________________________________
<p>Scientific Network : <a href="http://www.falconfly.de/network.htm">36200 MHz «» 8204 MB «» 815.0 GB</a> </p>

Windows 95/98/ME do not have all of the process controls that NT/2K/XP/2003 have, and therefore it is quite possible that on these machines, there is no solution available. I have read reports of this in the S@H boards for several weeks now for earlier versions of BOINC.
<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=13"><img border="0" height="80" src="http://13.cpdn.sig.boinc.dk?188"></a>

Well, since LHC and SETI work perfectly fine under Win98, I see no reason the CPDN should do different (?)

After all, it's just the one single point of actually pausing the Client, not more.

IMHO, it is a Bug inherent to the current CPDN Client, LHC or SETI's process control within BOINC works perfect.

Since the Sources for the SETI Client is available, it might be in order to have a look at the Code they used for process control, and compare it with the defective CPDN counterpart to help quickly spot the error.

...because noone can tell me it is impossible to pause/suspend a Program under Win98, it's just a basic issue (spotting the error itself in the Code might not be, of course).

Apart from the resulting CPDN performance which is reduced by about 50%, the machine works flawless with BOINC. Thanks to the powerful AthlonXP 3000+ and the generous CPDN WorkUnit Deadline, it remains useful even at that performance level ;)
(it just "could" be a whole lot quicker on all Projects now, that's it, and that's the only effect of the Bug)
___________________________________________
<p>Scientific Network : <a href="http://www.falconfly.de/network.htm">36200 MHz «» 8204 MB «» 815.0 GB</a> </p>

Yes, you have to go into the boinc general preference settings and set the Leave apps in memory preference to NO. Otherwise on windows 98 boinc is unable to fully stop a unit. As one of the few boinc alpha testers with a 98 machine I've noticed this to be a problem for the entire 4.x series. The unit does stop processing, but the CPU time keeps increasing. The result of this is just that the unit requests an excessively large amount of credit.

&gt; Yes, you have to go into the boinc general preference settings and set the
&gt; Leave apps in memory preference to NO.
I have applied this setting in the past with other BOINC Projects and believe i have lost processing time.<br>Is this possible?<br>
<a href="http://www.inetsonic.com/piskun/stats.htm"><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=151"></a>

From the FAQ


http://climateapps2.oucs.ox.ac.uk/cpdnboinc/quick_faq.php#a


"If you are running Windows 98, ME or Windows NT4:

Stick with classic version, as BOINC cannot yet cope with Windows 98, ME or NT4.
Unfortunately it does not look as if this can be easily solved.
The number of machines involved looks small so it is not considered to be worth the effort.
Sorry folks."
<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=lhcah&amp;userid=3724"><img border="0" height="80" src="http://3724.lhcah.sig.boinc.dk?188"></a>

&gt; From the FAQ
&gt;
&gt; http://climateapps2.oucs.ox.ac.uk/cpdnboinc/quick_faq.php#a
&gt;
&gt; "If you are running Windows 98, ME or Windows NT4:
&gt;
&gt; Stick with classic version, as BOINC cannot yet cope with Windows 98, ME or
&gt; NT4.
&gt; Unfortunately it does not look as if this can be easily solved.
&gt; The number of machines involved looks small so it is not considered to be
&gt; worth the effort.
&gt; Sorry folks."<br><br>
<b><i>Maybe FAQ should be updated as per this recent e-mail:</i></b><br>
Sent: Thursday, September 30, 2004 3:17 AM
Subject: <b>climateprediction.net Win98/ME/NT4 problems fixed</b>

Hello,

Forgive the email, but in case you were trying to run climateprediction.net
on a Win98, WinME, or Win NT4 machine and did not have success,
we have identified and corrected a bug in the Fortran code that was
preventing older operating systems from being able to run.

So if you "Reset", or "Detach" and re-"Attach" to climateprediction.net using your
CPDN/BOINC authenticator key, it should get this new version (4.04) of
Windows code that will run fine on 98/ME/NT4.

If you have any problems, or if you want to report success (or failure)
please do not hesitate to contact myself or Tolu Aina (tolu@atm.ox.ac.uk)
via email.

Sincerely,

Carl Christensen
Chief Software Architect
climateprediction.net<br>

<a href="http://www.inetsonic.com/piskun/stats.htm"><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=151"></a>

Hm, thanks for the Info, but by now, I have found 3 other Systems (Linux) exhibiting the same Bug; this time affecting SETI, which creeps along this way...

The mentioned Win98 machine was not attached to CPDN until the release of the recent V4.04 Client.

Not 100% sure, but it might not be Win98, nor the CPDN Client itself.

Either my Monitoring the Network with BOINCview 0.82d (via GUI_RPC) interferes with the BOINC Client, or BOINC V4.09 is showing this bug after running several days without being shutdown.

Might indeed also be the "Leave Application in memory" Option, but disabling it (technically causing the Clients to quit running after timeslicing to another Project) didn't make any difference it seems. I'll have to monitor this over a longer period of time with the Option disabled for all Projects now.
(would be odd, however, since until about last week, everything definitely ran perfectly normal with it enabled)

For that reason, I've posted this into the SETI/BOINC Forums as a general Bug report; maybe something turns out.

I'd hate to reset the Project(s), just to find out the error might not be fixed by it.
___________________________________________
<p>Scientific Network : <a href="http://www.falconfly.de/network.htm">36200 MHz «» 8204 MB «» 815.0 GB</a> </p>

&gt; <b><i>Maybe FAQ should be updated as per this recent e-mail:</i></b><br>

E-mail didn't come my way. Maybe because I run XP. Thanks for the update.

<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=lhcah&amp;userid=3724"><img border="0" height="80" src="http://3724.lhcah.sig.boinc.dk?188"></a>

thanks for reminding me, I did just update the FAQ to reflect that Win98/NT4/ME should be OK now.

&gt; &gt; <b><i>Maybe FAQ should be updated as per this recent e-mail:</i></b><br>
&gt;
&gt; E-mail didn't come my way. Maybe because I run XP.<br>A lot of people didn't receive this e-mail. I think many isp's block it as spam.<br>I had to use my alternate yahoo address before i started to receive.<br>
<a href="http://www.inetsonic.com/piskun/stats.htm"><img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=151"></a>