Thread sleep/wakeup (#31510 updated) #32029
Conversation
vtjnash
force-pushed
the
jn/kp/threadsleep
branch
from
2fa1969 to
604668c
Compare
base/task.jl
Outdated
| @@ -184,6 +184,7 @@ end | |||
|
|
|||
| # NOTE: you can only wait for scheduled tasks | |||
| function wait(t::Task) | |||
| t === current_task() && error("deadlock detected") | |||
There was a problem hiding this comment.
| t === current_task() && error("deadlock detected") | |
| t === current_task() && error("Possible deadlock detected: Cannot wait on current task") |
There was a problem hiding this comment.
Since it's certain deadlock, saying "possible" seems odd
There was a problem hiding this comment.
I was thinking about user-error, where someone literally does wait(current_task()). But in the end I only care that it is clear which action is not allowed.
There was a problem hiding this comment.
Yes, I ran into this because I did foreach(wait, tasks), then corrupted a huge amount of memory (basically, I just dropped all GC roots) which put current_task into tasks.
In full assertion mode, we could even expand this to check that donenotify forms a DAG, this just constitutes a simple (e.g. inexpensive) subset of the general case. I think other systems do something similar to that, but on a timer?
vtjnash
force-pushed
the
jn/kp/threadsleep
branch
from
604668c to
78017b7
Compare
vtjnash
force-pushed
the
jn/kp/threadsleep
branch
from
b917c98 to
204c29a
Compare
JeffBezanson
force-pushed
the
jn/kp/threadsleep
branch
2 times, most recently
from
991ab79 to
0c6cac8
Compare
|
OK I think this is ready to merge. |
JeffBezanson
force-pushed
the
jn/kp/threadsleep
branch
from
0c6cac8 to
52d7c18
Compare
JeffBezanson
force-pushed
the
jn/kp/threadsleep
branch
from
52d7c18 to
93e3d28
Compare
This works towards integrating partr scheduler more closely with Julia.