BT 0.99

Started by Pepo, February 07, 2011, 08:07:03 PM

Previous topic - Next topic

0 Members and 2 Guests are viewing this topic.

fred

Quote from: Beyond on February 17, 2011, 04:07:48 PM
What Windows version are you using.
Win7-64.  Even tried a different mouse.  No change.  Also found that if I set The "ready to start" filter (combine) to off so that the number of enries is larger the same thing often happens in Tasks.  Tried cutting my history to 8 hours, it made no difference.
[/quote]
Do you only click or move the selection with the arrow keys.

Pepo

A great bunch of notices appeared in BoincTasks from PrimeGrid, like this:
QuotePrimeGrid: World Record Generalized Fermat Prime
On 8 Feb 2011 09:38:57 UTC, PrimeGrid’s PRPNet found the largest known Generalized Fermat Prime: 145310^262144+1 The mega prime is 1,353,265 digits long and enters Chris Caldwell's The Largest Known Primes Database ranked 1st for Generalized Fermat Primes and 23rd overall. The discovery was made by Ricky L Hubbard of the United States using an AMD PHENOM II X6 1090t @ 3.2GHz with 8GB RAM, running Windows 7. This computer took about 11 hours to probable prime (PRP) test with GenefX64 and 48 hours and 15 minutes to complete the primality test using pfgw64. Ricky is a member of the AMD Users team. For more details, please see the official announcement.
10.02.11 06:20 viac...

In BOINC Manager, they appeared a bit differently:
QuotePrimeGrid: World Record Generalized Fermat Prime
On 8 Feb 2011 09:38:57 UTC, PrimeGrid's PRPNet found the largest known Generalized Fermat Prime: 145310^262144+1

The mega prime is 1,353,265 digits long and enters Chris Caldwell's The Largest Known Primes Database ranked 1st for Generalized Fermat Primes and 23rd overall.

The discovery was made by Ricky L Hubbard of the United States using an AMD PHENOM II X6 1090t @ 3.2GHz with 8GB RAM, running Windows 7. This computer took about 11 hours to probable prime (PRP) test with GenefX64 and 48 hours and 15 minutes to complete the primality test using pfgw64. Ricky is a member of the AMD Users team.

For more details, please see the official announcement.
26.02.2011 5:08:20 . viac...

Original message can be seen on PrimeGrid's News archive page.

The BOINC Manager's version correctly applied the message's formatting, whereas BT incorrectly discarded all line breaks. Additionally the apostroph in "PrimeGrid's PRPNet" was incorrectly rendered as "PrimeGrid’s PRPNet".
OTOH, BT correctly reflected all posts' original dates and times, whereas BOINC Manager incorrectly applied "26.02.2011 5:08:20" on all newsfeed items - probably the moment my client or Manager noticed them?
Peter

fred

Quote from: Pepo on February 28, 2011, 08:50:08 PM
A great bunch of notices appeared in BoincTasks from PrimeGrid, like this:
Ok on the bug list, may need a conversion table, did see something like this in the BOINC code.

Pepo

Quote from: Pepo on February 28, 2011, 08:50:08 PM
A great bunch of notices appeared in BoincTasks from PrimeGrid, [...]

[...] BT correctly reflected all posts' original dates and times, whereas BOINC Manager incorrectly applied "26.02.2011 5:08:20" on all newsfeed items - probably the moment my client or Manager noticed them?
In the mean time I've updated BOINC from 6.12.12 to 6.12.15. Possibly as a subsequence of this (although, WHY AT ALL ???), the client discarded all notices. A half day later they all appeared again, unfortunately in exactly reversed order (not noticeable in BOINC Manager, as all are "01.03.2011 13:58:23" (I've already complained on mailing lists), but funny to see the reverse-ordered dates in BT :D)

Some common and useful ordering is probably to be agreed on...
Peter

Beyond

Personally think the whole BOINC notice fiasco is a mess.  Feels like spam.  It's going to get worse too as more projects start sending out irrelevant messages.

Pepo

I suspect we should observe it from the positive side - fun for participants  :D

(until it eventually gets sorted out by volunteers (and the patches eventually checked in))
Peter

Pepo

#21
Buggy still:

  • the tray icon's "GPU suspended" indicator is really still buggy... I'm using it multiple times daily and it often keeps solid red even if nothing is suspended (and BT also thinks so, according to its tray popup menu without any checkbox and countdown timer).
Peter

fred

Quote from: Beyond on March 01, 2011, 03:50:45 PM
Personally think the whole BOINC notice fiasco is a mess.  Feels like spam.  It's going to get worse too as more projects start sending out irrelevant messages.
A spam filter maybe.

Beyond

Quote from: fred on March 01, 2011, 04:31:18 PM
Quote from: Beyond on March 01, 2011, 03:50:45 PM
Personally think the whole BOINC notice fiasco is a mess.  Feels like spam.  It's going to get worse too as more projects start sending out irrelevant messages.
A spam filter maybe.
BOINC notices spam filter, there's a business opportunity.  Personally wish it could be completely turned off.  Every one of my machines recently got something like 11 messages from PrimeGrid about some drivel that I could find out if I wanted by visiting their home page.  Even though "Notice reminder interval" is set to 0, they still popup.  Irritating.

fred

Quote from: Pepo on February 28, 2011, 08:50:08 PM
A great bunch of notices appeared in BoincTasks from PrimeGrid, like this:
You may try this:

In the changing the notices_template.html

charset=__InsertCharSetCode__
charset=utf-8

This solves the problem on my computer.
You may have to restart BT

Pepo

Quote from: Beyond on March 01, 2011, 04:49:07 PM
Personally wish it could be completely turned off.  Every one of my machines recently got something like 11 messages from PrimeGrid about some drivel that I could find out if I wanted by visiting their home page.
If multiple machines, controlled by BT, get the same notice, then maybe at least BT could pick one particular piece and simply discard the remaining thousand or so (or actually a bit less) identical ones?

I know that the notices seem to be bound to hosts, but maybe some project news (so, not error messages and warnings, bound to particular hosts) could be considered as being common to all hosts, and be removed from particular host to some artificial category?
Peter

Pepo



Quote from: fred on March 01, 2011, 05:42:55 PM
Quote from: Pepo on February 28, 2011, 08:50:08 PM
A great bunch of notices appeared in BoincTasks from PrimeGrid, like this:
You may try this:
In the changing the notices_template.html

charset=__InsertCharSetCode__
charset=utf-8

This solves the problem on my computer.
You may have to restart BT
Wasn't "__InsertCharSetCode__" intended to insert the charset from my .btlang file? Or BT's Language settings?
Really. It solved the apostroph problem, but broke all localized texts ;D

Quote from: notices_template.html<!--
To create a user template that isn't overwritten by the installer....
Copy this file to notices.html in the same folder as the exe.

Search sequence:

1) notices.html is used.
2) if notices.html isn't found notices_template.html is used.
-->
I've also tried to copy the template into notices.html. It worked, just the file was not expected to be found around the exes, but in the same html folder as the template. Due to change the comment?

BTW, what about such subtle modification? --- C:/Program Files/eFMer/BoincTasks/html/notices_template.html ut III  1 23:20:20 2011
+++ C:/Program Files/eFMer/BoincTasks/html/notices.html ut III  1 23:35:58 2011
@@ -16,15 +16,16 @@

<!--
__InsertNoticeTemplateMoreBegin__
-<FONT COLOR="#0000ff" SIZE=3><a href="__InsertNoticeLink__" target="_new">__InsertNoticeMore__</a></FONT>
+<FONT COLOR="#0000ff" SIZE=1><a href="__InsertNoticeLink__" target="_new">__InsertNoticeMore__</a></FONT>
__InsertNoticeTemplateMoreEnd__
-->

<!--
__InsertNoticeTemplateBegin__
<FONT SIZE=3><B>__InsertNoticeTitle__</B></FONT><br>
+<FONT COLOR="#bbbbbb" SIZE=2>__InsertNoticeTimeCreated__</FONT><br>
<FONT SIZE=2>__InsertNoticeBody__</FONT><br>
-<FONT COLOR="#bbbbbb" SIZE=2>__InsertNoticeTimeCreated__</FONT>__InsertNoticeMoreLink__
+<FONT COLOR="#bbbbbb" SIZE=1>(__InsertNoticeTimeArrival__)</FONT>__InsertNoticeMoreLink__
<hr size="1" align="left" width="100%" color="#bbbbbb">
__InsertNoticeTemplateEnd__
-->

(At least I know I can use it personally, just have to check the template for updates.)
Peter

Pepo

#27
Today I've noticed two auto-aborted tasks. Corresponding messages:

10414 Renderfarm.fi 03.03.11  06:26 Aborting task ses0000001672frm0000000878prt00001_1: exceeded memory limit 2433.54MB > 2413.72MB
10467 Renderfarm.fi 03.03.11  06:32 Aborting task ses0000001672frm0000000890prt00001_2: exceeded memory limit 2449.61MB > 2413.72MB


I've also sampled a couple of failed results from my History, the upper two correspond with the aborts:

Application   Elapsed time   Completed   Reported Status
4.84 Blender 00:00:00 (00:00:00) 03.03.11  06:32 03.03.11  06:34 Aborted (-177)
4.84 Blender 00:00:00 (00:00:00) 03.03.11  06:26 03.03.11  06:28 Aborted (-177)
2.31 Data coll. 00:00:00 (00:00:00) 04.02.11  12:01 04.02.11  12:01 Aborted (194)
4.83 Blender 00:02:51 (00:02:01) 21.12.10  19:31 21.12.10  19:33 Reported: Computation error (-177,)
4.83 Blender 00:01:07 (00:00:18) 21.12.10  15:26 21.12.10  15:28 Reported: Computation error (-177,)
5.03 GARLI 00:00:00 (00:00:00) 26.11.10  14:05 26.11.10  14:38 Aborted by project
2.02 DNETC 00:00:00 (00:00:00) 26.11.10  12:22 26.11.10  12:25 Aborted (-1)
5.05 Astropulse 1+20:18:01 (1+11:34:45) 15.10.10  21:59 15.10.10  22:04 Aborted by user


My questions:

  • Can the "Aborted (-177)" status be directly translated to some particular (human-readable) abort reason? Or is 177 the application's own error code? The tasks' logs contain "Exit status   -177 (0xffffff4f)" + "<message> Maximum memory exceeded </message>" + "Exception caught: BOINC forcibly tells us to abort right now! Status: 0"
  • Is it possible, that "Computation error (-177,)" meant the same lack-of-memory reason? (OK, found "Exit status   -177 (0xffffff4f), Exception caught: BOINC forcibly tells us to abort right now! - Unhandled Exception Record - Reason: Out Of Memory (C++ Exception) (0xe06d7363) at address 0x000007FEFD4FAA7D" in the task's result.)
  • Can (194) and (-1) states be directly translated to some particular (human-readable) abort reasons? I remember for sure I've aborted the (locked) "2.31 Data collect" task (on 4.2.2011) manually from BT and reported about it on WUProp forum.

Funny these messages, in addition to "Exception caught: BOINC forcibly tells us to abort right now!" I've also found a "Exception caught: BOINC kindly asks us to exit", even multiple times in the same log  :D
Peter

Pepo

Quote from: Pepo on March 01, 2011, 10:53:11 PM
Quote from: fred on March 01, 2011, 05:42:55 PM
Quote from: Pepo on February 28, 2011, 08:50:08 PM
A great bunch of notices appeared in BoincTasks from PrimeGrid, like this:
You may try this:
In the changing the notices_template.html

charset=__InsertCharSetCode__
charset=utf-8

This solves the problem on my computer.
You may have to restart BT
Wasn't "__InsertCharSetCode__" intended to insert the charset from my .btlang file? Or BT's Language settings?
Really. It solved the apostroph problem, but broke all localized texts ;D
What about to convert localized parts of notices' text to UTF-8 prior to inserting them into final notices.html document? Then UTF-8 could be applied on the whole doc?
Peter

fred

Quote from: Pepo on March 03, 2011, 03:00:24 PM
What about to convert localized parts of notices' text to UTF-8 prior to inserting them into final notices.html document? Then UTF-8 could be applied on the whole doc?
That's what I'm doing now. But still, I don't know why PrimeGrid's uses UTF-8 at all. ;D
I now try to get the ASCII text to be translated to UTF-8. From the right code page that is.