|
21)
Message boards :
Number crunching :
app config
(Message 779)
Posted 21 Nov 2015 by  Ananas
Post: I know that it might be a bit far-fetched on a Linux machine but as I cannot see what's wrong ... If you created the file with windows notepad and then moved it to Linux, it might have a character set identifier in the first two characters of the text file, indicating that it has been saved as UTF8 If this is the case, it would confuse the XML parser of the core client and it wouldn't recognize the first <app_config> tag |
|
22)
Message boards :
News :
Longer work units
(Message 689)
Posted 2 Nov 2015 by Ananas
Post: Not sure if I've got the new WUs or not ... You currently have 2 types in progress, the ones with a one week deadline are the "medium new" ones, the ones with a two weeks deadline the "brandnew" ones. The one-week WUs have a somewhat longer runtime and a really bad estimated runtime (by mistake), the two-week ones run shorter but have a much higher estimated runtime. As your host probably has adjusted the correction factor to the one-week results, the two-week results will probably be listed with an estimated runtime of several days but will run less than one day. The one-week results have a better time to credits ratio than the two-week results. As soon as you have only brandnew (two-weeks) results left, the estimated runtime will slowly adjust to more realistic values I guess (but I'm not so experienced with the new estimate mimic of the 7.x core client so I might be wrong there). |
|
23)
Message boards :
Number crunching :
Too many errors (may have bug)
(Message 687)
Posted 2 Nov 2015 by Ananas
Post: I don't want to be misunderstood. My point was not that I think that the application is faulty in any way. I just think that the error quota should be increased, as a lot of hosts seem to have trouble with things that are not connected to the application itself. With the current setting, the risk is (too) high that a successful result doesn't make it into the science database |
|
24)
Message boards :
News :
Android application
(Message 682)
Posted 1 Nov 2015 by Ananas
Post: JagDoc (he posted his host links above) uses a client version that reports several peak space ressources used at runtime. p.s.: The BOINC guys probably have added attributes to the result table for this, but of course those are only filled when the client reports them. I never have seen those values in older projects, so it iseems to require both client and server from a fairly new version. p.p.s.: hmmm ... I proposed something like this long ago at CPDN in order to help developers set proper rsc_something_bound values, maybe someone picked it up :-) |
|
25)
Message boards :
News :
Android application
(Message 663)
Posted 30 Oct 2015 by Ananas
Post: Disk space calculator - really old but I guess that BOINC still works like that : http://oct31.org/boincspace/ p.s. : WAIT ... isn't that the one for violation of RSC_DISK_BOUND? In this case it would not be a host setting but a WU setting. You could check the setting in client_state.xml (best use a copy of that file, not the original one) Find the Universe workunit : <app_name>universe-xray_sources_v3</app_name> (the version might vary) then a few lines below that : <rsc_disk_bound>somefloatingpointnumber</rsc_disk_bound> |
|
26)
Message boards :
Number crunching :
Potentially unfair credit awarded
(Message 650)
Posted 28 Oct 2015 by Ananas
Post: Result nr. / WU nr. 6136679 / 2844232 ... 6112779 / 2832285 ... different hosts the long running one is a T2400 on Windows the short one a Core2Quad Q9450 on Linux The much faster CPU running the way better optimized application version, so it is not too surprising that it has a much better time to credits ratio |
|
27)
Message boards :
Number crunching :
Too many errors (may have bug)
(Message 647)
Posted 27 Oct 2015 by Ananas
Post: An old BOINC bug, it sends out results that wouldn't ever have had a chance to be valid, caused by the limits : max # of error/total/success tasks 3, 10, 6 Especially with the long running ones, the error limit is too low. I have one with wingmen errors : Not started by deadline - canceled Aborted by user Error while computing (an unreliable box that has ~1/3rd errors) Completed, can't validate (mine *sigh*) Error while computing (ERR_RESULT_START, result not even started) Error while downloading (ERR_RESULT_DOWNLOAD) None of the errors is caused by the result itself but still it is invalid http://universeathome.pl/universe/workunit.php?wuid=2850372 The other one looks very similar. http://universeathome.pl/universe/workunit.php?wuid=2850252 I aborted three now (_5 and _6) because the risk to waste a lot of CPU time is much too high with those settings. p.s.: The _6 results are always doomed, if a workunit had nothing but error results so far. BOINC still sends one out (this is the ancient BOINC server side bug) p.p.s.: A second really old BOINC bug is that hosts with only 2% valid results still can have the full quota, reported a lot of times, but in Berkeley no one cares. |
|
28)
Message boards :
Number crunching :
Something is messy on the project prefs page
(Message 636)
Posted 21 Oct 2015 by Ananas
Post: For the same reason (rudimentary XML parser) the XMLs contain more linefeeds than necessary (from the XML syntax view of things). Please don't optimize those as it might affect client_state.xml and mangle the contents of all projects ;-) |
|
29)
Message boards :
Number crunching :
Something is messy on the project prefs page
(Message 634)
Posted 21 Oct 2015 by Ananas
Post: Besides the layout bug (which still exists but imo. isn't too critical), the venues don't work at all. The problem isn't easy to find, it's a server side problem in the work scheduler - so it might be a Berkeley bug. sched_reply....xml contains invalid venue tags. All projects : <venue name="school"> Universe : <venue name='school'> Note the single quotes, the BOINC core client has only a rudimentary XML parser which seems not to be able to handle anything but double quotes. Other than that, the sched_reply looks fine, this is the only irregular feature I could find. The problem must be fairly new as I remember to have used a separate resource_share for this venue and it used to be working. Now it ignores the setting and uses the one from "default venue" p.s.: Here we have the parser part ... a reference string is built : sprintf(buf, "<venue name=\"%s\">", venue_name); and then compared to the (trimmed) read buffer. As the reference string contains double quotes, the comparison isn't tolerant towards the other type of quotes. The bug probably sits in prefs.inc, function global_prefs_make_xml(), this is from an older version : $xml = $xml."<venue name=\"school\"> that's where the double quotes should be generated. |
|
30)
Message boards :
News :
Longer work units
(Message 620)
Posted 16 Oct 2015 by Ananas
Post: 1,538.46 will be final.... Is it exactly that or more like 1538,4615384615384615384615384615 ? |
|
31)
Message boards :
News :
Longer work units
(Message 615)
Posted 15 Oct 2015 by Ananas
Post: It doesn't matter for me but some have complained about the upload size in the past so it's probably worth to mention. The upload size seems to have been reduced, when I saw one that was just uploading, it was only about half the size of the previous results. |
|
32)
Message boards :
Number crunching :
Linux doesn't get any WU ?
(Message 608)
Posted 14 Oct 2015 by Ananas
Post: Linux app is only available for x64 (check "Application" link above) |
|
33)
Message boards :
News :
Longer work units
(Message 604)
Posted 14 Oct 2015 by Ananas
Post: OT : The WU names somehow bring up memories to Roelof Engelbrecht, who passed away November 16, 2004 - crunchers of SETI Classic (pre-BOINC) sure still remember him and probably his weblog about his fight against cancer, that he lost on that day. |
|
34)
Message boards :
News :
Longer work units
(Message 603)
Posted 14 Oct 2015 by Ananas
Post: Runtime on my Xeon L5520 will be ~24 hours, I received 91(!) workunits yesterday - very hard deadline if this isn't the only project p.s.: cache setting = store 0.75 + additional 0.25 so I guess you didn't increase RSC_FPOPS_EST |
|
35)
Message boards :
Number crunching :
Something is messy on the project prefs page
(Message 584)
Posted 5 Oct 2015 by Ananas
Post: The quoting is OK now, the table structure is still messy.png On the image you can see where inner and outer table are closed and (in bold red) the table row / table data that are not included in any table and below the code the rendering outcome. |
|
36)
Message boards :
Number crunching :
Something is messy on the project prefs page
(Message 581)
Posted 4 Oct 2015 by Ananas
Post: Thanks for reporting the issue. No, the one I'm using for "school" is still messy - different messy than before but still messy. Now the variable text is cluttered below the default venue, still no HTML table for any view but the default. It can still be visualized and modified by using the alternative view (venues side by side) though, that one has no trouble. p.s.: maybe quoting the HREF contents would help. Plus the <tr>+<td> stuff would better be within a table, not outside ;-) The table is closed just before this (unquoted) href : <a href=add_venue.php?venue=home&subset=project&tnow=....>Add separate pref ... p.s.: I'm aware that dealing with generated HTML code with inline PHP statements can be somewhat tricky sometimes. If you're using an editor that can fold the code (like Arachnophilia), it might help to hide the PHP tags in order to see what's in the wrong place. Or use the source view of Mozilla, the damaged parts will be displayed in red. |
|
37)
Message boards :
Number crunching :
False positiv virus alarm? Trojan.GenericKD.2649361 found.
(Message 580)
Posted 4 Oct 2015 by Ananas
Post: Those download errors I mentioned above - is it possible that the files are already rejected on a proxy server? In this case there would be hardly a chance to fix it, if the scanner on the proxy doesn't support whitelists. |
|
38)
Message boards :
Number crunching :
Something is messy on the project prefs page
(Message 566)
Posted 28 Sep 2015 by Ananas
Post: http://universeathome.pl/universe/prefs.php?subset=project shows the "edit" links in a weird way : First some links Add separate preferences for home then the default prefs with edit link, then the school prefs without edit link and then the page footer Mozilla 31.0 p.s.: this must have happened within the last 1 or 2 days, it did work before |
|
39)
Message boards :
Number crunching :
False positiv virus alarm? Trojan.GenericKD.2649361 found.
(Message 560)
Posted 27 Sep 2015 by Ananas
Post: Some hosts (not mine) get this message : <message> app_version download error: couldn't get input files: <file_xfer_error> <file_name>universe-xray_sources_v3_2_windows_intelx86.exe</file_name> <error_code>-224 (permanent HTTP error)</error_code> <error_message>permanent HTTP error</error_message> </file_xfer_error> </message> The file is not missing in the download directory, so this might be a result of this false virus alert I guess (?) p.s.: the code tag doesn't do what it is supposed to, same for pre |
|
40)
Message boards :
Number crunching :
Cancelled by server
(Message 546)
Posted 19 Sep 2015 by Ananas
Post: When one host doesn't return his result within the deadline, another result is sent out to compensate this. If the delayed host returns his result after the deadline, but before the replacement result has been startet (edit : and if it is valid), the server side scheduler asks the host that received the replacement result to trash that one. You can recognize replacement results. With a quorum of two the two original results have a suffix _0 and _1. The first replacement result has _2, if another replacement is required _3 and so on. |
(English)
(Polski)