Threading plus multiprocessing plus cv2 error

[John O'Hagan]

On Sat, 29 Aug 2020 13:01:12 -0400
Dennis Lee Bieber <wlfraed at ix.netcom.com> wrote:

> On Sat, 29 Aug 2020 18:24:10 +1000, John O'Hagan
> <research at johnohagan.com> declaimed the following:
> 
> >There's no error without the sleep(1), nor if the Process is started
> >before the Thread, nor if two Processes are used instead, nor if two
> >Threads are used instead. IOW the error only occurs if a Thread is
> >started first, and a Process is started a little later.
> >
> >Any ideas what might be causing the error?
> >  
> 
> 	Under Linux, multiprocessing creates processes using fork().
> That means that, for some fraction of time, you have TWO processes
> sharing the same thread and all that entails (if it doesn't overlay
> the forked process with a new executable, they are sharing the thread
> until the thread exits).
> 
> https://stackoverflow.com/questions/54466572/how-to-properly-multithread-in-opencv-in-2019
> (which points to)
> https://answers.opencv.org/question/32415/thread-safe/?answer=32452#post-id-32452
> """
> The library itself is thread safe in that you can have multiple calls
> into the library at the same time, however the data is not always
> thread safe. """
> 
> 	The sleep(1), when compounded with the overhead of starting
> the thread, and then starting the process, likely means the thread
> had exited before the process actually is started. Try replacing the
> sleep(2) in the work code with something like sleep(30) -- I
> hypothesize that you'll get the same error condition even with the
> sleep(1) in place.
> 
> 

Thanks for the reply. 

You're right, the error also happens with a longer sleep, or no sleep
at all, inside the function. That sleep is only there in the example to
keep the windows open long enough to see the images if they are
successfully displayed.

I could well be wrong as I'm not fully across multiprocessing, but I
think the Stackoverflow question and answer you linked above relate to a
different situation, with multithreaded cv2 operations on shared image
data. 

In my example, AFAIK (which is not very far) it shouldn't matter whether
the new process is sharing the thread, or whether the thread has
exited, because the thread and the process aren't using the same data.
Or (as is quite likely) am I misunderstanding your point?

Cheers

John