A month ago, I upgraded to Windows 10 Professional (32 bit). I am running BOINC 7.6.22.

When I reboot my computer, those tasks with partial progress before the reboot, after the reboot these tasks are restarted, they either:
- Restart with 0% (the beginning) progress OR
- They terminate with an "Error while Computing".

This is happening for all 10 projects that I'm participating in. It makes it impossible to run some projects with long run times, without a reboot, such as Climate Prediction.

Tasks restarting from the beginning points to a project science application that does no checkpointing. WCG does have a couple of those. There's nothing you can do about that, other than not reboot or run the tasks to the end first.

As for the error while computing, under the taskID it gives a more in-depth error message that may give a clue as to what happened.

BOINC does like it much when it is writing the task state to disk while at that moment the computer reboots, because it can't gracefully exit at that moment. Easy workaround here: exit BOINC before you allow the computer to reboot.

Which projects is your BOINC attached to that have these problems? Links to examples?

---

Task / Work unit / Status / Run time / CPU time / Application

107920919 / 45016363 / Error while computing / 2,040.02 / 1,812.05 / Period Search Application v102.10 (avx)

107920932 / 45016369 / Error while computing / 4,698.94 / 2,943.64 / Period Search Application v102.10 (avx)

107919654 / 45016374 / Error while computing / 12,823.10 / 10,112.39 / Period Search Application v102.10 (avx)

4813847527 / 2104003498 / Error while computing / 2,941.01 / 100.11 / SETI@home v8 v8.00 (opencl_intel_gpu_sah) windows_intelx86

4811404274 / 2102839735 / Error while computing / 3,610.72 / 141.77 / SETI@home v8 v8.00 (opencl_intel_gpu_sah) windows_intelx86

123660466 / 122974906 / Error while computing / 31.53 / 24.16 / CPUMiner (Campaign #13) v0.01

697966806 / 473892807 / Error while computing / 1,962.00 / 1,203.56 / Sierpinski Problem ESP/PSP/SoB (Sieve) v1.12

With links to example tasks, I mean, for example the Seti task you're pointing to:
Right-click on the TaskID number, from the drop-down menu choose "Copy link location" or "Copy shortcut", then to post that link here in an answer window, right click in the answer window and choose "Paste", or left-click in the answer window and press CTRL+V on the keyboard.

Like so: https://setiathome.berkeley.edu/result.php?resultid=4813847527

What you've now done is impossible for me to follow, because I really don't know which projects have which applications. Which is also why I asked for the projects you've got added and have this problem with.

---

By the way, the error you have at Seti is: (unknown error) - exit code -1073741205 (0xc000026b)

In Windows, error 0xc000026b means "The device has no drivers installed, or the drivers are configured incorrectly".

I see that you're using the Intel GPU built into the Intel CPU you have. It requires correct drivers with correct OpenCL to do work at Seti (and possibly other projects). The latest driver -that Windows 10 always wants to install- is not necessarily the best driver for the GPU, because it needs to be compatible with the projects and their science application.

You can do this by telling Windows 10 not to search for and install driver updates (see http://winsupersite.com/windows-10/stop-automatic-driver-updates-windows-10 for how to do that).
Then you have to go to the Intel support website and download the drivers from there, like these. Download them, save them to disk or your desktop, then (double-)click them to install them.

---

http://asteroidsathome.net/boinc/result.php?resultid=107920944

http://asteroidsathome.net/boinc/result.php?resultid=107920932

http://asteroidsathome.net/boinc/result.php?resultid=107919654

http://setiathome.berkeley.edu/result.php?resultid=4813847527

http://setiathome.berkeley.edu/result.php?resultid=4811404380

http://www.primegrid.com/result.php?resultid=697966806

http://www.bitcoinutopia.net/bitcoinutopia/result.php?resultid=123660466

As I suspected, in all projects bar Bitcoinc Utopia you err on 0xc000026b, so follow the advice I gave above on adjusting the videocard drivers. Or stop using the Intel GPU on these projects.

The exit code 195 (0xc3) error on Bitcoin Utopia means that their wrapper failed. Probably again due to driver trouble.

---

Actually, Windows error 0xC000026B is defined in https://msdn.microsoft.com/en-us/library/cc704588.aspx as

0xC000026B
STATUS_DLL_INIT_FAILED_LOGOFF
{DLL Initialization Failed}

The application failed to initialize because the window station is shutting down.

The problem has just been reported at https://github.com/BOINC/boinc/issues/1553, together with a possible solution - use boinctray to monitor windows messaging (it does that), and create a mutex to signal the client (which doesn't).

Tasks restarting from the beginning points to a project science application that does no checkpointing. WCG does have a couple of those. There's nothing you can do about that, other than not reboot or run the tasks to the end first.

Seen you write this twice now in a short while and wonder what the source of this story is, as all sciences at WCG checkpoint, either by force [Ebola single job tasks checkpoint 8 times] to Clean Energy, checkpointing 5 times, after each sub-job, which can take a long time if running on crawlers, but they all checkpoint, either programmed or use the 'at most' setting in the client prefs/location profile. If you see not, then maybe bring this via your internal channels to the WCG techs.

Edit: For users... BOINCTasks has a wonderful setting where you can tell a job(s) to pause when reaching checkpoint. Then it's a case of suspending of long-apart-checkpoint tasks 'ready to start' and you're set to do a planned boot without loosing too much crunching time on resume.

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Seen you write this twice now in a short while and wonder what the source of this story is

At the time of me writing that, March 26 2016, the source had been a thorough search through the WCG forums and reading several threads on the subject from top to bottom two weeks earlier: I'd checked there on the 12th of March, and at the time more specifically for my answer to the Android app. But saw at that time that there were a couple of other apps that didn't checkpoint or didn't checkpoint correctly as well.

A lot can change in nine and a half weeks in development.

At least in the case of the Android answer I asked the person to go ask on the WCG forums. And mind, at the time of writing the Zika app wasn't even in Frage.

Shall I come over to the WCG forums and resurrect some old threads in which you gave advice from nine weeks ago that's no longer correct? You could've answered this in fifteen different ways, all not attacking me. What's up, Rob? Want to tell me something?

---

'im Frage' is the pseudo-German term ;). 15 different ways to answer, they all checkpoint, and have been since shortly after production release. The only one that didn't was Ebola, which took some work, but that was early 2015 when updated. Telling people there aren't is a tech issue, hence the question... Android some report loosing all progress [for we don't know why, which is not the same as not checkpointing, for which there unfortunately does not exist a tick box to log them in the app], but this thread is on Windows.

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

'im Frage' is the pseudo-German term ;).

Sentences end with a full stop before the smiley, not the smiley and then the full stop.

but this thread is on Windows.

Well great. You've answered that it's all fixed now.
Yippee.

---

WCG techs were working on smaller chunk CEP2 jobs in beta [a 'know what you get into' opt-in science], but have pulled the plug for now and moved the research last night to intermittent. The first of 5 steps could still take > 18 hours in latest beta, which is the cut-off time. Snap-shot checkpointing 2.5GB output files, which is the size the models could gain, is just not a good idea... crunching while using and the mouse lagging 2 streets behind.

:).

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Back to the 'crashes due to reboot' problem. I think I see what may have gone wrong. From https://msdn.microsoft.com/en-us/library/cc231199.aspx:

0x0000026B
ERROR_INVALID_HW_PROFILE
The specified hardware profile configuration is invalid.

That sounds close enough to "The device has no drivers installed, or the drivers are configured incorrectly" that it perhaps implies a simple mis-reading between

0000026B
C000026B

I do wish Microsoft would pay more attention to their typography.