Problems interrupting pmap with Ctrl-C

[amitmurthy]

Unable to interrupt a pmap call with 2 or more workers

julia> addprocs(2)
julia> pmap(x-> while true; sleep(2000); end, 1:4)
^C^C^C^C^C^C^C^C^C^C^C^C

I added a debug statement to the sigint handler as well as the try-catch block in pmap. Looks like while the sigint handler is called everytime, only one the async tasks in pmap (for the first Ctrl-C) is being interrupted.

julia> addprocs(2)
julia> pmap(x-> while true; sleep(2000); end, 1:4)
^Cjl_defer_signal 0, jl_signal_pending 0, sig 2
InterruptException()
^Cjl_defer_signal 0, jl_signal_pending 0, sig 2
^Cjl_defer_signal 0, jl_signal_pending 0, sig 2
^Cjl_defer_signal 0, jl_signal_pending 0, sig 2
....

Are all the tasks interrupted upon a Ctrl-C? If so, then it could be a REPL bug too.

If not, the "main" task, which in the case of pmap is the REPL waiting on a @sync block should be interrupted, instead of a @async task chosen at random.

[vtjnash]

^C interrupts the currently executing task, which may be the last worker to block

@JeffBezanson maybe InterruptException should be propagated to the root task, if the current task is in the uv_run scheduler?

[StefanKarpinski]

Bump. Would be nice to have a fix for this. Interrupting a long-running distributed task is pretty common.

[simonster]

Bump. This is really annoying.

[ihnorton]

Related (dupe?) #6283

[tkelman]

@amitmurthy have recent refactorings improved the situation here at all?

[bjarthur]

possibly fixed. i get the following error on OS X for the latest master, and sth similar on linux:

julia> addprocs(2)
2-element Array{Int64,1}:
 2
 3

julia> pmap(x-> while true; sleep(2000); end, 1:4)
^C^C^C^Cfatal: error thrown and no exception handler available.
InterruptException()
jl_run_once at /Users/arthurb/src/julia/src/jl_uv.c:141
process_events at ./libuv.jl:82
wait at ./event.jl:100
task_done_hook at ./task.jl:161
unknown function (ip: 0x31ee9ecd9)
jl_call_method_internal at /Users/arthurb/src/julia/src/./julia_internal.h:93
jl_apply at /Users/arthurb/src/julia/src/./julia.h:1384
^C^C
atexit hook threw an error: InterruptException()^Cjl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
^Cjl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF
jl_uv_writecb() ERROR: bad file descriptor EBADF

[bjarthur]

hold on. while on OS X it does return me to the command line, on linux it does not. rather, it keeps repeating the error message over and over.