Closed mdhe1248 closed 7 years ago
I've seen this too (edit: to clarify, in other code, not registration). I think it's a Julia codegen bug. Does it happen if you don't use multiple workers and just run everything in a single process?
It does not happen if I don't use multicore.
On Sep 12, 2017 4:42 AM, "Tim Holy" notifications@github.com wrote:
I've seen this too. I think it's a Julia codegen bug. Does it happen if you don't use multiple workers and just run everything in a single process?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/HolyLab/BlockRegistrationScheduler/issues/41#issuecomment-328800625, or mute the thread https://github.com/notifications/unsubscribe-auth/AMvfFQ5lmon8r3H1ld8wsy8A_Xqzc3NEks5shlIAgaJpZM4PSchI .
Now it is working without the error. I am not sure why it happened before. It may happen again if I forcibly quit registration process or quit by itself due to incorrect inputs/parameters. Until I get the error, I will close it.
I guess this is a frustrating error. I don't think it is easy to repeat the error in other circumstances. Nevertheless, recently I often observed "ReadOnlyMemoryError" here and there. There should be some problem. I hope someone provides insights on this matter.
I connected cannon from my laptop (ubuntu 16.04) using ssh -X. If I use wpids = [1], this error does not occur. When I run multi process, worker 2 always gives the error. If I don't use worker2, then worker3 gives the error.
This might be solved if I update packages. However, the newest package versions throw another error during BlockRegistration.