GR.jl
GR.jl copied to clipboard
jheinen
Cannot interrupt `plot` in a loop
julia> using GR; for i in 1:10
plot(1:10000000)
end
^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C
julia> versioninfo()
Julia Version 1.1.0
Commit 80516ca202 (2019-01-21 21:24 UTC)
Platform Info:
OS: Linux (x86_64-pc-linux-gnu)
CPU: Intel(R) Core(TM) i7-6820HQ CPU @ 2.70GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-6.0.1 (ORCJIT, skylake)
(v1.1) pkg> st GR
Status `~/.julia/environments/v1.1/Project.toml`
[28b8d3ca] GR v0.39.0
I'm not sure whether this should/can be handled within GR!?
The following snippet would work:
using GR
for i = 1:10
try
plot(1:100000)
println(i)
sleep(1)
catch InterruptException
println("Interrupt")
break
end
end
julia> using GR
julia> for i = 1:10
try
plot(1:100000)
println(i)
sleep(1)
catch InterruptException
println("Interrupt")
break
end
end
1
2
^C3
^C^C4
^C^C^C^C5
^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C6
^C^C7
8
9
10
Unfortunately, it doesn't quite work.
Which platform are you using?
I am using Arch Linux.
Did you invoke Julia using julia -i ... ?
No, I didn't.
This happens on Fedora Linux as well. Have to just kill -9 my Julia scripts that use GR.
It seems to me this is a problem with Julia itself. It's odd that even with a large sleep statement outside of GR, you still can't get Julia to respond to a SIGINT. How could GR be responsible for something that is happening outside of GR? Although, none of that changes the fact that GR is the only package I've had this problem with.
I just tried several OSs: macOS and Windows 10 work fine - on Ubuntu 18, it doesn't work. The problem seems to be specific to Linux. I'm currently out of office, but it's on my radar ...
I have the same problem (on Linux) with this script:
for i = 1:10
try
io = open("foo.txt", "w")
println(io, 1:10000)
close(io)
println(i)
sleep(1)
catch InterruptException
println("Interrupt")
break
end
end
... which means, that it's not related to GR (IMO).
@jheinen I was able to get stuck in the loop you just posted. Just to be clear, this is the loop I get stuck in:
for i = 1:10
try
io = open("foo.txt", "w")
println(io, 1:10000)
close(io)
println(i)
sleep(1)
catch InterruptException
println("Interrupt")
break
end
end
However, this only happens if I have imported GR and done a plot. If I do not import GR, then I can interrupt the loop you posted just fine. If I do import and use GR, then I can no longer break from the loop you posted. I am on Linux / Fedora 31.
It seems to me that this issue is exclusive to GR. Yet, your examples makes us ask, how does GR cause problems in a loop which is not even calling GR? It could be an issue with Julia, but every example of this problem I have seen involves GR.
MacOS, Julia 1.6.2, GR 0.58.1. Simply GR.clearws() totally kills Ctrl+C even for loops without GR in them, that would interrupt otherwise. The problem doesn't exist in 0.55.0.
$ julia
_
_ _ _(_)_ | Documentation: https://docs.julialang.org
(_) | (_) (_) |
_ _ _| |_ __ _ | Type "?" for help, "]?" for Pkg help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 1.6.2 (2021-07-14)
_/ |\__'_|_|_|\__'_| | Official https://julialang.org/ release
|__/ |
julia> for i=1:1000000 for j=1:1000000 i^j end end
^C^C^C^C^CWARNING: Force throwing a SIGINT
ERROR: ^C^C^CInterruptException:
Stacktrace:
[1] >>
@ ./int.jl:455 [inlined]
[2] >>
@ ./int.jl:462 [inlined]
[3] power_by_squaring(x_::Int64, p::Int64)
@ Base ./intfuncs.jl:276
[4] ^
@ ./intfuncs.jl:290 [inlined]
[5] top-level scope
@ ./REPL[1]:1
julia> ^C
julia> using GR
julia> GR.clearws()
julia> for i=1:1000000 for j=1:1000000 i^j end end
^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C^C
You will have to disable the internal exit handler in GR's underlying GKS, e.g.
env GKS_NO_EXIT_HANDLER=1 julia
In many graphics applications, an exit handler in the graphics subsystem makes perfect sense. In your case, however, it must be switched off.
Unfortunately, it doesn't seem to fix the problem for me
➜ env GKS_NO_EXIT_HANDLER=1 julia
_
_ _ _(_)_ | Documentation: https://docs.julialang.org
(_) | (_) (_) |
_ _ _| |_ __ _ | Type "?" for help, "]?" for Pkg help.
| | | | | | |/ _` | |
| | |_| | | | (_| | | Version 1.6.2-pre.0 (2021-04-23)
_/ |\__'_|_|_|\__'_| | release-1.6/dd122918ce (fork: 333 commits, 265 days)
|__/ |
julia> using GR
julia> for i = 1:10
try
plot(1:100000)
println(i)
sleep(1)
catch InterruptException
println("Interrupt")
break
end
end
1
2
^C3
^C^C4
^C^C5
^C^C^C6
^C^C^C7
^C^C8
^C^C^C^C^C^C^C^C^C^C9
^C^C^C^C^C^C^C^C^C^C10
^C^C^C^C^C
julia> versioninfo()
Julia Version 1.6.2-pre.0
Commit dd122918ce (2021-04-23 21:21 UTC)
Platform Info:
OS: macOS (x86_64-apple-darwin20.3.0)
CPU: Intel(R) Core(TM) i7-1068NG7 CPU @ 2.30GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-11.0.1 (ORCJIT, icelake-client)
Environment:
JULIA_NUM_THREADS = 4
JULIA_PKG_DEVDIR = /Users/scheme/src/julia
JULIA_PKG_SERVER = https://neuralsim.juliahub.com
I can't reproduce the behaviour with our binaries. Could you please try to build GR using ENV["JULIA_GR_PROVIDER"] = "GR"? It seems to me, that the problem is related the BinaryBuilder run-time.
No, doesn't help. I even tried the following:
sigsetmask = t -> ccall(:sigsetmask, Cint, (Cint,), t)
sigblock = t -> ccall(:sigblock, Cint, (Cint,), t)
mask = sigblock(0)
using GR
GR.clearws()
grmask = sigblock(0)
using Printf
@printf("orig mask: %x, after GR: %x, GR w/o SIGTERM block: %x\n", mask, grmask, grmask & ~(1<<14))
sigsetmask(mask)
According to it, GR disables SIGTERM (signal 15, bit 14) receipt, so I tried re-enabling it, but it doesn't help. In any case Control+C should send SIGINT… but that one (signal 2, bit 1) is blocked with or without GR. Weird… Also just sending SIGINT to julia directly causes it to crash. So Julia must have some weird nonstandard ^C processing.
I can't reproduce the behaviour with our binaries. Could you please try to build GR using
ENV["JULIA_GR_PROVIDER"] = "GR"? It seems to me, that the problem is related theBinaryBuilderrun-time.
How is that done?
My next attempt would be to study tcgetattr to see if there are some terminal Ctrl+C settings, but that's more work as C structs would be needed… maybe one day.
Based on the above issue I reported against Julia, I also tested sending SIGINT directly instead of with Ctrl+C to Julia while running the final loop in my earlier comment https://github.com/jheinen/GR.jl/issues/211#issuecomment-909376759: for i in $(seq 1 10); do kill -SIGINT $JULIAPID; done. Nothing, the loop won't terminate. So the issue is not about the terminal configuration as suggested in my previous post. It has to be something about how SIGINT is being processed internally in Julia after GR is loaded.
Also, if I just kill -SIGINT $JULIAPID after loading GR (using GR; GR.clearws()) but keeping the REPL idle, nothing happens. However, if GR is not loaded, Julia crashes (https://github.com/JuliaLang/julia/issues/42072). So GR somehow totally disables SIGINT handling, but according to my other comment https://github.com/jheinen/GR.jl/issues/211#issuecomment-909414800, this does not seem to be directly using signal masks.
