[QUOTE=Prime95;138707]This is the same bug S000113 just reported. It is now fixed.[/QUOTE]I wasn't sure, cause I noticed this ~11 hours after you reported fixing it. However it's possible my bad value was in there from before your fix and it just didn't misbehave until I tried looking at the Test|Status. I'll assume it really is fixed unless it shows up again.

[QUOTE=Prime95;138663]The save file should still be in your directory. You can manually add the exponent back to your worktodo.txt file.

Does prime.log show any errors or weird messages that ight shed a clue as to why prime95 started on new exponents?[/QUOTE]Did the former with worktodo.add

It appears that when I executed the manual com (see [URL="http://www.mersenneforum.org/showpost.php?p=138660&postcount=69"]this post[/URL], this was the real machine B), it appears that v5 told the machine, "That doesn't belong to you". So the machine stopped the expo. I don't have the prime.log available at this time. But, did look and there was a message.

I've recently bought a new Core 2 Quad and am running Vista 64. I ran the Torture Test without errors for the past two days and am now trying to test exponents (1 DC and 3 first times which are currently in P-1 stage). After about 30 minutes or so (time varies) Windows says that Prime95 has stopped responding and closes the program. Is it possible that I allocated too much memory (I assigned 4GB of 8GB to Prime95)?

[QUOTE=jketch;138788](1 DC and 3 first times which are currently in P-1 stage). [/QUOTE] The 3 P-1's are likely a problem. Convert 2 to TF's and After the first expo has cleared the P-1, change one to first time, then repeat when that has cleared the p-1 phase. After that, you should not see 3 P-1's all at once again.

The Mac link needs a .org added to the domain and the binary fails with a bus error. Mac OS X 10.4.11

Version 25.6 with v4 server
 

Two days ago I downloaded v25.6 and started using it with the old server. After finishing the first P-1 the worktodo.txt looks like[CODE][Worker #1]
Test=000000000000000000000000024F5EB1,38756017,68,1

[Worker #2]
Test=000000000000000000000000024F5F33,38756147,68,0

[Worker #3]
Test=000000000000000000000000025145DD,38880733,68,0

[Worker #4]
Test=000000000000000000000000025152E3,38884067,74,1
[/CODE](After I got the assignments, I restarted, using only one core.) However, the primenet report looks like[CODE]38756017 75 5757543 1.8 35.2 60.2 06-Aug-08 07:23 04-Aug-08 11:52 nu 2405 v19/v20
38756147 * 68 1.8 34.2 59.2 04-Aug-08 11:53 nu 2405 v19/v20
38880733 * 68 1.8 34.2 59.2 04-Aug-08 11:54 nu 2405 v19/v20
38884067 75 1.8 34.2 59.2 04-Aug-08 11:54 nu 2405 v19/v20
[/CODE]I.e. as if trial factoring was done up to 74 bits. Also look at the assignment I got for thread #4. That was trial factored to 74 bits by someone else. Or maybe just a P-1 like this one was done.

The output was[CODE][Work thread Aug 6 09:21] Stage 2 GCD complete. Time: 189150.7 ms.
[Work thread Aug 6 09:21] M38756017 completed P-1, B1=465000, B2=12903750, E=6, Wd2: EA6B2758
[Comm thread Aug 6 09:21] Sending result to server: UID: patrik/nu, M38756017 completed P-1, B1=465000, B2=12903750, E=6, Wd2: EA6B2758
[Comm thread Aug 6 09:21]
[Comm thread Aug 6 09:21] URL: http://mersenne.org/cgi-bin/pnHttp.exe?ar&4&88&4&23293&64746&patrik&[I](removed in post)[/I]&nu&38756017&0&74.5
[Work thread Aug 6 09:21] Resuming primality test of M38756017 using FFT length 2048K
[Work thread Aug 6 09:21] Iteration: 350626 / 38756017 [0.90%].
[Comm thread Aug 6 09:21] RESPONSE:
[Comm thread Aug 6 09:21] pnResult=0
[Comm thread Aug 6 09:21] URL: http://mersenne.org/cgi-bin/pnHttp.exe?rm&4&133&4&12218&32450&patrik&[I](removed in post)[/I]&nu&UID:+patrik/nu,+M38756017+completed+P-1,+B1=465000,+B2=12903750,+E=6,+Wd2:+EA6B2758%0A
[Comm thread Aug 6 09:21] RESPONSE:
[Comm thread Aug 6 09:21] pnResult=0
[Comm thread Aug 6 09:21] Done communicating with server.
[/CODE]

Two more things. (Haven't read through this thread to see if it has already been reported.)

I first edited worktodo.txt without stopping mprime first. Then it crashed. V24 didn't behave like that. (I did some benchmarking right after getting the assignments, so I edited worktodo.txt to skip the P-1. After I was finished I edited them again to do the P-1. This explains the 0.90% done above.)

When I first registered I didn't type in my User Name. It defaulted to " cycles from this program.".

The version I have used is[CODE]-r-xr-xr-x 1 owner group 731193 Jan 30 13:04 mprime256.tar.gz
[/CODE]but I just now realize that there is a 64 bit version as well. My system is[CODE]pjoh@nu:~/v256_mprime> uname -a
Linux nu 2.6.18.2-34-default #1 SMP Mon Nov 27 11:46:27 UTC 2006 x86_64 x86_64 x86_64 GNU/Linux
pjoh@nu:~/v256_mprime> more /etc/SuSE-release
openSUSE 10.2 (X86-64)
VERSION = 10.2[/CODE]

[QUOTE=phy1729;138832]The Mac link needs a .org added to the domain and the binary fails with a bus error. Mac OS X 10.4.11[/QUOTE]

Fixed the link. Thanks.

I've only got 10.5 systems here - so I think the binaries will require 10.5 to run. Try going to the 25.5 thread and download that binary. I think it was built when I had 10.4 installed.

I'll add it to my to-do llist to investigate how to build a binary that works on both 10.4 and 10.5.

[QUOTE=patrik;138842]I.e. as if trial factoring was done up to 74 bits.[/QUOTE]

This is a harmless side effect of v25.6. The code that translates v5 messages to send to the v4 server doesn't have enough info to send an accurate bit level (all it knows is that trial factoring is complete). So, it tells the v4 server that the exponent is factored to 74 bits knowing this will prevent the v4 server from handing the exponent out for further trial factoring.

The master database is updated by a different means. It uses the text messages sent to the v4 server. So, if you look up the exponent on the v5 server you'll see the true bit level.

[quote=Prime95;138852]Fixed the link. Thanks.

I've only got 10.5 systems here - so I think the binaries will require 10.5 to run. Try going to the 25.5 thread and download that binary. I think it was built when I had 10.4 installed.

I'll add it to my to-do llist to investigate how to build a binary that works on both 10.4 and 10.5.[/quote]

25.5 works on 10.4.11

There are two things in Prime95 I sorely miss after switching over from StressPrime2004 (Orthos). #1 is beeping on errors, #2 is stop stress testing on error. Can these two simple features be added to the next version of P95 please for us stress testers out there?

George, you should consider adding a bug tracking system to the GIMPS website. This would help identify bugs more quickly.