There's no recent info on the new script that was being worked on.
So once a week it is.

Andy asked the other day about how the credit granted by the script on the testing site compares with that here, in preparation for introducing the new script so there is movement. However I would suggest that the time for breath holding is not yet! Unless others have replied privately to Andy, only two of us have given feedback so far and I would be looking to make sure there is something approaching a consensus before starting the new script.

Since this is the tread where we discussed this the last time that we had a Credit Export problem ... I will bring it back here.

It appears The Credit Stat's are not being Exported on a regular basis again. Here are the current time stamps of the data last exported,

Last update user XML 2023-05-01 19:00:17 UTC (34 days 07:19:56 old)
Last update host XML 2023-04-30 18:50:17 UTC (35 days 07:29:56 old)
Last update team XML 2023-05-01 19:00:17 UTC (34 days 07:19:56 old)

About 35 Days ago.

Some one needs to bump the Script.

Thanks
Bill F

Some one needs to bump the Script.

But not till after the issue with credit doubling has been resolved.

Edit: Email sent to Andy.

If some extra credit is awarded accidentally, is that a bad thing really? There have been plenty of times where credit are lost accidentally, right?

If some extra credit is awarded accidentally, is that a bad thing really? There have been plenty of times where credit are lost accidentally, right?

It’s the volume of the extra credits that’s the problem, that and the extreme RAC figures generated that do not appear to be coming down.

OK, we occasionally loose a few thousand credits in a glitch but gaining a few million credits is a bad thing.

Well it has been more days and while we don't know for sure if the credit script is ever being run one thing we do know is that Stat's are not being exported. Here is a snapshot of the export files with time stamps.

Last update user XML 2023-05-01 19:00:17 UTC (51 days 05:06:45 old)
Last update host XML 2023-04-30 18:50:17 UTC (52 days 05:16:45 old)
Last update team XML 2023-05-01 19:00:17 UTC (51 days 05:06:45 old)

Dave can you inquire ?

Dave can you inquire ?

I could but there has been a flurry of emails about credit and how much should be granted for each of the new model types between Richard and Andy the last being 15:17 UTC on Tuesday. We are nearing a decision on that. Once that is sorted, I will chase it if needed.

Update: yesterday, I received a substantial tranche of sample data for local analysis - around 80,000 tasks and 1.2 million trickles, covering the last six months of the project (specifically, including the whole of the IFS production run). So far, I've

• converted it into a suitable format for desktop processing
  
• imported it into a desktop database, and performed basic checks (created a valid index, converted dates to Windows format)
  
• queried for my own tasks, and compared the results with a data-scrape I'd already done from this site. They matched.
  

So, the extract from the 'result' table that I received matches what you can see for your tasks here - that's reassuring.

Today, I intend to do the same thing for the trickle table (had y'all noticed that trickles for IFS tasks were added to the database from 15:00 on 9 May, and became visible for your tasks?). That was the day before we noticed the massive jump in our total credits - methinks that the two dates must be connected.

I'm not sure yet whether I've got the right details to reproduce the "credit from trickles" that the old (2009) script produced. I'll be looking to see if I can recreate the numbers for either total credit or RAC, or even possibly both. If I can, I'll let you know.

But so far, I've not seen one single line of actual computer code relating to the 'new' way of calculating credit - just some over-simplified hand-waving narrative descriptions. One of which was utterly, completely, bonkers. So it's a bit frustrating.

Dave... Thanks that makes sense let the dust settle a little

Update 2: I've now done the analysis for the trickle table too. With one tiny (trivial) exception, it matches the task values.

But the overall increase in my user total has gone up by a much larger jump - over 90%. That where I'll need to focus next. I've just had an email from Andy with a link: I need a break, but I'll look at it after lunch.

Hmmm. This isn't going so well. It isn't a script - it's a full C++ program, written with loads of subroutine calls in the usual inside-out spaghetti order, with the main entry point at the bottom of the last file you look at. Anybody on this board got any experience of turning that into a linear flow chart, or other possible solutions? It's in a public Git repository, but it's got calls to external libraries and (I think) external closed-source programs.

Credit export still to be sorted but, the new batch (those that don't crash anyway) have zips and credit showing on the task pages.

Credit export still to be sorted but, the new batch (those that don't crash anyway) have zips and credit showing on the task pages.

Are you in a position to monitor and log the extra credit you receive on a task-by-task basis, and compare it with the total credit shown daily or weekly on your user account? (*)

My analysis so far is suggesting that the individual task figures are sensible compared with what we saw before, but the user account total is greater than the sum of its parts. There's also a big question over RAC, but it turns out that the data dump I received is missing one crucial field: I've asked for that, but it's not going to be made available until Tuesday.

And I still have that dreadful C++ tangle to unravel.

* It's possible (probable?) that user totals are still not being updated, while stats export is still disabled. But they are threatening to re-start the exports 'soon', ready or not, accurate or not. It would be really, really, helpful to have some precision task logs ready to compare with the aggregate figures when they appear.

It's possible (probable?) that user totals are still not being updated, while stats export is still disabled.

My user total and average have not changed since 30th May despite the tasks being granted credit.

I have two instances of BOINC running under WINE. One is in a VM. The one running in the host OS has stopped responding to both mouse and keyboard. I am a bit reluctant to kill it and risk losing the tasks. I could mitigate the risk by using boinccmd to suspend the tasks before restarting I suppose. Once the current tasks are complete I need to try and see if i can reproduce the problem and if so file a bug report.

The BOINC instance in the VM is a slightly older iteration. The problem one I downloaded yesterday.

Edit: used boinccmd to suspend and then shut down client; xkill to shut down manager, restarted and all running fine.

From event log
24/06/2023 09:48:25 | climateprediction.net | [http] [ID#10] Info: Closing connection 7
24/06/2023 09:48:25 | climateprediction.net | [http] [ID#10] Info: schannel: shutting down SSL/TLS connection with www.google.com port 443
24/06/2023 09:48:25 | climateprediction.net | [http] [ID#10] Info: Hostname in DNS cache was stale, zapped
24/06/2023 09:48:25 | climateprediction.net | [http] [ID#10] Info: Trying 141.223.16.156:80...
24/06/2023 09:48:26 | climateprediction.net | [http] [ID#10] Info: connect to 141.223.16.156 port 80 failed: Connection refused
24/06/2023 09:48:26 | climateprediction.net | [http] [ID#10] Info: Failed to connect to upload7.cpdn.org port 80 after 1000 ms: Couldn't connect to server
24/06/2023 09:48:26 | climateprediction.net | [http] [ID#10] Info: Closing connection 9
24/06/2023 09:48:26 | climateprediction.net | [http] HTTP error: Couldn't connect to server
24/06/2023 09:48:26 | climateprediction.net | Temporarily failed upload of wah2_eas25_a3oo_201511_25_994_012220214_2_r14589202_1.zip: connect() failed
24/06/2023 09:48:26 | climateprediction.net | Backing off 00:14:24 on upload of wah2_eas25_a3oo_201511_25_994_012220214_2_r14589202_1.zip
24/06/2023 09:48:27 | | Project communication failed: attempting access to reference site
24/06/2023 09:48:27 | | [http] HTTP_OP::init_get(): https://www.google.com/
24/06/2023 09:48:28 | | [http] [ID#0] Info: Trying 172.217.16.228:443...
24/06/2023 09:48:28 | | [http] [ID#0] Info: Connected to www.google.com (172.217.16.228) port 443 (#10)
24/06/2023 09:48:28 | | [http] [ID#0] Info: schannel: disabled automatic use of client certificate

I see today, my RAC has taken a big hit which suggests the excess due to OIFS tasks getting inflated credit may have been resolved.

I see today, my RAC has taken a big hit which suggests the excess due to OIFS tasks getting inflated credit may have been resolved.

I'll have to wait until Andy sends me those final data fields on Tuesday. If they arrive early in the day, I can check them before I have to set out for Northumberland: otherwise, it'll have to wait until I can set up my laptop in the hotel room.

But it's a good sign - I would expect the RAC decay to kick in as soon as new work starts to flow. It's just the quantum I want to check.

I see you have a total credit of 15,539,788 - how does that compare with the figure before the server move?

Edit - my 33,188,860 total and 1,559,478 RAC are still the inflated figures, but I haven't had the energy to chase this latest batch release. Nothing to prod the server to touch my account.

I see you have a total credit of 15,539,788 - how does that compare with the figure before the server move?

Total credit was about 8,000,000 on 25th May. RAC was about 68,000 on 23rd, yesterday dropped to just under 5,000 then a slight increase today. On second thoughts, maybe that is just the script for RAC running again.

Hmmm. This isn't going so well. It isn't a script - it's a full C++ program, written with loads of subroutine calls in the usual inside-out spaghetti order, with the main entry point at the bottom of the last file you look at. Anybody on this board got any experience of turning that into a linear flow chart, or other possible solutions? It's in a public Git repository, but it's got calls to external libraries and (I think) external closed-source programs.

Richard, when faced with a new code I can't follow, I normally fire it up in the debugger and step through the code to see what it does. Or, put in various print statements to see.

Thanks for the suggestions. My problem is that all my serious development work has been done under Windows, more recently using VBA for Access (office suite automation against a SQL server database - that got quite tasty after a while), and - starting under Windows 3, so 30 years ago - with the first commercial release of Visual Basic.

I have done some minor C++ patches for BOINC, using one or other of the free community versions of Microsoft Visual Studio, but found I have to renew my licence - but didn't have time to follow through before starting this trip. I have started to create a flowchart using Notepad++ - that has code-aware colour coding and the ability to search multiple files for function calls and definitions at the same time. But it's a slow process.

This morning I looked at the final data file, received from Andy while I was driving north. But I discover that all the IFS task types have a zero entry in the (previously missing) trickle_timestep field. So my hopes of comparing the old and new script/program calculations, including RAC, have been dashed.

I'll need to get my thinking cap out next week. The biggest question is: "How is RAC calculated for IFS?" - because that has to work for the old model types too.

There are tree analysis tools out there, though I don't know for Windows. But data-dependent branching will make your static analysis more time consuming.

If the calculation for the credit resides in one function and only one function, maybe that's all you need to look at and understand the inputs? Could then create a standalone version of the same calculation in python or a language of your choice?

As you know I've kept well away from the credit side of things - I have enough to do. Maybe someone in the boinc 'team' might know better?