PRPNet 3.0.0 Released - Page 9

rogue · 2010-02-17, 22:14

Quote:

Originally Posted by Xentar

With the fix, the access violation didn't occurr again.
But today in the morning, there was another error:

The error message is in german, I can try to translate it:
[2010-02-17 04:43:45 GMT] Connect to database failed: [Microsoft][ODBC Driver Manager] Fehler beim SQLAllocHandle-Aufruf für Treiber auf SQL_HANDLE_ENV, native code=0
means something like
[2010-02-17 04:43:45 GMT] Connect to database failed: [Microsoft][ODBC Driver Manager] Error while calling SQLAllocHandle of driver at (or in?) SQL_HANDLE_ENV, native code=0

Seems, like the server lost the database connection, but I don't know why. I was sleeping at 4:40am

When a client connects, it gets its own unique connection handle to the database, so this problem would have been localized to that client.

Did this only happen with one client or did other clients get the same problem?

It seems that a memory leak is possible, but I would have expected the application to crash if that happened. I will be adding memory leak detection code (Windows only) in the next release. If there is a memory leak, it will be caught.

Xentar · 2010-02-17, 22:26

Sorry, I had to give more details.

When this happened, just two clients were active, but no one got new tests, until I restarted the server at 6:50am.
Sadly, I disabled debug mode a few days ago. But from the log file I see, that the server was running since [2010-02-14 17:17:39 GMT]
So, a memory leak is possible. I can keep an eye on it. While I am writing this text, the memory usage of prpserver.exe is 75MB.

Restarting the server now, and it goes down to... 4 MB.

Edit 2:
By the way: Is it normal, that the socket number always increases? The last one before this crash was 193k.

Quote:

[2010-02-17 04:42:05 GMT] Socket 193908: 325816*19^36314+1 sent to Email: <> User: xentar Client: 01
[2010-02-17 04:42:46 GMT] Socket 193920: 341956*19^36314+1 sent to Email: <> User: xentar Client: 02

rogue · 2010-02-17, 22:50

Quote:

Originally Posted by Xentar

Sorry, I had to give more details.

When this happened, just two clients were active, but no one got new tests, until I restarted the server at 6:50am.
Sadly, I disabled debug mode a few days ago. But from the log file I see, that the server was running since [2010-02-14 17:17:39 GMT]
So, a memory leak is possible. I can keep an eye on it. While I am writing this text, the memory usage of prpserver.exe is 75MB.

Restarting the server now, and it goes down to... 4 MB.

Edit 2:
By the way: Is it normal, that the socket number always increases? The last one before this crash was 193k.

That doesn't make any sense. A memory leak is the only explanation I can think of that would cause the problem. I won't be able to do much about it until the weekend.

Xentar · 2010-02-18, 22:04

Just for your information: After 24 hours, the memory usage of prpserver.exe is 110MB. Something's wrong here.

Anything else, how I can help?

rogue · 2010-02-19, 02:02

Quote:

Originally Posted by Xentar

Just for your information: After 24 hours, the memory usage of prpserver.exe is 110MB. Something's wrong here.

Anything else, how I can help?

You can't do much right now. I hope to track down the memory leaks this weekend. I will release a 3.2.1 if I can nail the big ones. That doesn't mean that all of the other issues have been solved, just some of them.

Mini-Geek · 2010-02-20, 11:26

I'm using PRPnet 3.2.0 on Windows XP with server type set to "1 - Sierpinski/Riesel". I added 192K-200K from Sierp 16. I think the server mode might've been 4 when I added the candidates (which, as I'll detail later, is the only mode I've tried where it still produces the right page). The server is handing out work fine, but the server_stats.html page is very odd. The code looks like this: (I tried copying the table, but that didn't work)

Code:

<html><head><title>PRPNet 3.2.0 Server Statistics - my PRPnet</title>
<style type="Text/css"></style></head><body>
<table frame=box align=center border=1>
<tr><td align=center colspan=10>Sierpinski base 16</tr>
<tr><td>Form<td>Total Candidates<td>Min N<td>Max N
<td>Count Tested<td>Count DC'd<td>Count Untested
<td>Leading Edge<td>PRPs/ Primes<td>Lowest Prime
</tr>
<tr>
<td align=center>ÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌ*16^n+1<td align=right> 1<td align=right>194207<td align=right>194207
<td align=right>0<td align=right>0<td align=right>1
<td align=right>0<td align=right>0<td align=center>none</tr>
<tr>
<td align=center>ÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌ*16^n+1<td align=right> 1<td align=right>194995<td align=right>194995
<td align=right>0<td align=right>0<td align=right>1
<td align=right>0<td align=right>0<td align=center>none</tr>
<tr>
<td align=center>ÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌ*16^n+1<td align=right> 1<td align=right>195023<td align=right>195023
<td align=right>0<td align=right>0<td align=right>1
<td align=right>0<td align=right>0<td align=center>none</tr>
<tr>
<td align=center>ÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌ*16^n+1<td align=right> 1<td align=right>195113<td align=right>195113
<td align=right>0<td align=right>0<td align=right>1
<td align=right>0<td align=right>0<td align=center>none</tr>

(Might it not be a coincidence that the hex for Ì is CC? I don't know, just stands out to me having both digits the same...)
The n's don't seem to be in any particular order, but are grouped together very roughly based on their size (you can see how the first four are about 195K).
Having server type set to "3 - Fixed b, k, and c, grouped by n" produces a similar page (same sort of nonsense garble, and the n's are arranged differently, but with the same general pattern of being roughly by size) but with "-858993460" (hex FFFF FFFF CCCC CCCC) instead of "ÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌÌ".
Setting it to "4 - Fixed b, n, and c, grouped by k" makes the page I'd expect that to show (with k*16^192002+1 at the top and going down from there).

Any idea what's up?

rogue · 2010-02-20, 13:36

Quote:

Originally Posted by Mini-Geek

I'm using PRPnet 3.2.0 on Windows XP with server type set to "1 - Sierpinski/Riesel". I added 192K-200K from Sierp 16. I think the server mode might've been 4 when I added the candidates (which, as I'll detail later, is the only mode I've tried where it still produces the right page). The server is handing out work fine, but the server_stats.html page is very odd. The code looks like this: (I tried copying the table, but that didn't work)

Any idea what's up?

Thanks for the bug report. I see the problem and will fix it in the next release.

mdettweiler · 2010-02-20, 18:03

Quote:

Originally Posted by Mini-Geek

I'm using PRPnet 3.2.0 on Windows XP with server type set to "1 - Sierpinski/Riesel". I added 192K-200K from Sierp 16.

Somewhat off-topic: when running a power-of-2 base through PRPnet, I'd recommend using a script (I've got a Perl script I can send you if you like) to convert the sieve file to base 2 first. That way, PRPnet will know to assign the job to LLR, thus doing an LLR or Proth test and producing such a residue; otherwise, it will assign the job to PFGW which will do a PRP test. The test may well end up taking about the same time in the long run, but since in general the residues we've collected for those bases are LLR/Proth instead of PRP, it would be less messy for future double-checking if we don't have to sort out which individual ranges are which type.

Mini-Geek · 2010-02-20, 18:29

PRPnet automatically sees that 16 is a power of 2 and passes it to LLR.

By the way, why does the PRPnet server say this when starting up?
"0 PRPs and Primes have been copied to the UserPrimes table"
And is it supposed to always say it (not just once or something)?

mdettweiler · 2010-02-20, 18:34

Quote:

Originally Posted by Mini-Geek

PRPnet automatically sees that base 16 is a power of 2 and passes it to LLR.

It does? Oh, right, I forgot...that was added sometime after I last ran base 16 work.

I think the last time I did that was on version 2.2 or something like that, and it passed base 16 to PFGW.

rogue · 2010-02-20, 18:56

Quote:

Originally Posted by Mini-Geek

PRPnet automatically sees that 16 is a power of 2 and passes it to LLR.

By the way, why does the PRPnet server say this when starting up?
"0 PRPs and Primes have been copied to the UserPrimes table"
And is it supposed to always say it (not just once or something)?

This is from a piece of code related to the upgrade that introduced the UserPrimes table. If that table is empty, then PRPNet thinks that it needs to be populated. Since there are no primes, then 0 got loaded into the table. This is continue to happen until a prime is found or until I remove the code that does the insert. In any case, there is no harm related to this message.

2010-02-20, 18:29	#97
Mini-Geek Account Deleted "Tim Sorbera" Aug 2006 San Antonio, TX USA 17·251 Posts	PRPnet automatically sees that 16 is a power of 2 and passes it to LLR. By the way, why does the PRPnet server say this when starting up? "0 PRPs and Primes have been copied to the UserPrimes table" And is it supposed to always say it (not just once or something)? Last fiddled with by Mini-Geek on 2010-02-20 at 18:33

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PRPNet 5.4.3 Released	rogue	Software	178	2021-06-24 11:56
PSP goes prpnet	ltd	Prime Sierpinski Project	86	2012-06-06 02:30
PRPNet 4.0.0 Released	rogue	Software	84	2011-11-16 21:20
PRPNet 4.0.1 Released	Joe O	Sierpinski/Riesel Base 5	1	2010-10-22 20:11
PRPNet released!	rogue	Conjectures 'R Us	250	2009-12-27 21:29

2010-02-18, 22:04	#92
Xentar Sep 2006 BB₁₆ Posts	Just for your information: After 24 hours, the memory usage of prpserver.exe is 110MB. Something's wrong here. Anything else, how I can help?