Hi there,

apparently there is a bug in the client (used by Menu Bar): it corrupts remaining work units after uploading results. I saw this four or five times. The pattern is:
- The client downloads couple of units (three in my example - I set exchange for 2 days),
- after the first unit is completed, it uploads it's results,
- then it DOES NOTHING but deferring communication with the project and increasing wait timer. The ~/Library/Application Support/Boinc Data/projects/setiathome.berkeley.edu/ folder contains the remaining units, but ~Library/Application Support/Boinc Data/slots/1/ is EMPTY (slot 1 because slot 0 is einstein on my machine).

The only way to get ride of this is to reattach to the project loosing remaining work units. When I go to RESULTS statistics, I can see the unfinished units as I PROGRESS. But they're already gone.

Who can fix that?

ALX

I don't know the answer to the question, but, does the client_state.xml file show the particulars of the remaining work units? If so, what does it show under state? It should show either a 0 or 1.

---

> ... what does it show under state? It should show either a 0 or 1.

Hmm, I don't have "bad state" file copy handy. I have to wait till it happens again:)
I'll let you know once I have it.

ALX

It happened again! Finished work unit and DOES NOTHING. Not even upload :-L

The last work unit 02ja05aa.4529.8624.353398.62_0 result:

name>02ja05aa.4529.8624.353398.62_0
final_cpu_time>19012.700000
exit_status>0
state>2
wu_name>02ja05aa.4529.8624.353398.62
report_deadline>1113502196.000000
file_ref>
file_name>02ja05aa.4529.8624.353398.62_0_0
open_name>result.sah

The work unit file is renamed to 02ja05aa.4529.8624.353398.62_0_0 and is 16kB instead of 356kB.

Active taks:

project_master_url>http://setiathome.berkeley.edu
result_name>02ja05aa.4529.8624.353398.62_0
active_task_state>0
app_version_num>402
slot>1
scheduler_state>1
checkpoint_cpu_time>19012.700000
fraction_done>0.822884
current_cpu_time>19107.190000
vm_bytes>0.000000


What can I do about it?

All files with underscore in their names in the ./projects/seti.berkeley.edu directory are the result files. Once the WU is finished, Seti tries to upload the result almost immediately. If Seti cannot contact the server for some reason (either your computer is not connected 24/7 or if the Berkeley server is down), Seti defers uploading the file for a few minutes, and later progressively for a few hours or days (It is easier to notice this in the command-line version) until it is able to talk to the server.

BTW, what kind of connection do you have - dialup or always online (cable, broadband, DSL, T1, etc.)? Also, connections during mornings US time (late morning or early afternoon in Poland) almost always goes through as opposed to connections during evenings US time.

---

I use DSL connection always online.

So _0 files are the results. What is _0_0 then?
The client just stopped to do anything while having one result and at least one work unit ready for computation.
I had to reattach to the project again.

I set the prefs to communicate twice a day instead once per two days. Let's see if this helps. I still have two work units files abandoned. Can I merge them somehow with the new works?

This is the seti only problem - einstein project works perfectly. I still have a strong impression there is a bug in the seti client. I did not have any problems with it's previous version. I saw other people complaining about exactly the same :-((

ALX

> So _0 files are the results. What is _0_0 then?

I'm not exactly sure, but I believe the first number after the underscore plus 1 refers to the number of computers the work unit has been sent to and the second 0 just denotes that the file is a result file.

I'm afraid I cannot think of anything else that might cause your problem. Maybe someone else has an idea?

---

OK, thanks for helping :-)

I switched to 3rd party compilation of the S@H by Team MacNN:

http://members.dslextreme.com/~readerforum/forum_team/boinc.html

It's based on older client and it's optimized for G4/5 with AltiVec - I'll check if the problem remains.

ALX

The S@H client by Team MacNN works fine!

ALX