Boinc 8.2.4 Windows Bug?
Posted: Sat Jul 26, 2025 6:20 am
There is a bug in the latest boinc under windows, a task that has a stalled upload can prevent all other projects being updated on completion, the other projects upload ok but are then prevented from reporting completion and hence will not download new work. Also manual commands to update projects are ignored. Restarting the boinc-client temporarily clears the log jam but while the upload is stalled it will happen again. Resetting the project with the stalled upload works but will lose credits from the relevant project and might need one more restart cycle.
I suspect the timed cycle for attempting to re-upload the stalled task is halting the updates, although to prove that would take days of faffing around and you need a task that has a stalled upload (goofygrid was the one in my case but denis and ithena are projects that also commonly end up in that state but all of them, are intermittent projects).
It might be a unique set of circumstances, goofygrid was still responding but doing so incorrectly (no start flag), if a server is simply down then it might not occur as that may have proper timeout handling. For the moment I've disabled goofygrid on Windows but the problem is most probably not limited to goofygrid.
I've not reported this bug, I spent about 5 hours getting as far as I did, luckily having three computers exhibiting the problem saved (or wasted?) a lot of time, I don't want to spend more time on diagnostics and I don't understand enough about boinc or windows to understand any interrogation on github.
The problem does not occur on Linux where I am much more comfortable debugging.
I suspect the timed cycle for attempting to re-upload the stalled task is halting the updates, although to prove that would take days of faffing around and you need a task that has a stalled upload (goofygrid was the one in my case but denis and ithena are projects that also commonly end up in that state but all of them, are intermittent projects).
It might be a unique set of circumstances, goofygrid was still responding but doing so incorrectly (no start flag), if a server is simply down then it might not occur as that may have proper timeout handling. For the moment I've disabled goofygrid on Windows but the problem is most probably not limited to goofygrid.
I've not reported this bug, I spent about 5 hours getting as far as I did, luckily having three computers exhibiting the problem saved (or wasted?) a lot of time, I don't want to spend more time on diagnostics and I don't understand enough about boinc or windows to understand any interrogation on github.
The problem does not occur on Linux where I am much more comfortable debugging.