issues with Kernel#select

[Tim Pease]

Waking up a thread that is waiting in Kernel#select does not appear to =
work in ruby 1.9 Can someone please confirm that this is the case. Is =
this the intended behavior, or is this a bug?

cat a.rb

require 'socket'

pair =3D Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t =3D Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join

ruby --version

ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]

ruby a.rb

Thread is about to exit ...

ruby1.9 --version

ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]

ruby1.9 a.rb #=3D> hangs forever!!!!

Any thoughts out there ???

Blessings,
TwP=

 

[Eric Wong]

Waking up a thread that is waiting in Kernel#select does not appear to work in ruby 1.9 Can someone please confirm that this is the case. Is this the intended behavior, or is this a bug?

cat a.rb

require 'socket'

pair = Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t = Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join

ruby --version

ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]

ruby a.rb

Thread is about to exit ...

ruby1.9 --version

ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]

ruby1.9 a.rb #=> hangs forever!!!!

Any thoughts out there ???

Hi Tim,

1.9 is actually doing what I expect it to do, that is waiting
indefinitely because the timeout argument for select is nil. I might
consider the 1.8.7 behavior a bug, but then again the underlying
select(2) syscall is allowed to have spurious wakeups. So even without
a timeout argument, select may always return even when nothing is
readable.

 

[Tim Pease]

to work in ruby 1.9 Can someone please confirm that this is the case. =
Is this the intended behavior, or is this a bug?

=20
=20

cat a.rb

require 'socket'
=20
pair =3D Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t =3D Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join
=20

ruby --version

ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]
=20

ruby a.rb

Thread is about to exit ...
=20

ruby1.9 --version

ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]
=20

ruby1.9 a.rb #=3D> hangs forever!!!!

=20
=20
=20
Any thoughts out there ???

=20
Hi Tim,
=20
1.9 is actually doing what I expect it to do, that is waiting
indefinitely because the timeout argument for select is nil. I might
consider the 1.8.7 behavior a bug, but then again the underlying
select(2) syscall is allowed to have spurious wakeups. So even = without
a timeout argument, select may always return even when nothing is
readable.

Thanks for the answer. My suspicion is that the Ruby 1.8 green threads =
can be woken up, but the Ruby 1.9 system threads "do the right thing"; =
hence, the discrepancy in observed behavior. Good to know that it's not =
a bug in 1.9.

/me goes off to rework code

Blessings,
TwP=

 

[Eric Wong]

Waking up a thread that is waiting in Kernel#select does not appear to work in ruby 1.9 Can someone please confirm that this is the case. Is this the intended behavior, or is this a bug?


cat a.rb
require 'socket'

pair = Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t = Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join

ruby --version
ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]

ruby a.rb
Thread is about to exit ...

ruby1.9 --version
ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]

ruby1.9 a.rb #=> hangs forever!!!!



Any thoughts out there ???

Hi Tim,

1.9 is actually doing what I expect it to do, that is waiting
indefinitely because the timeout argument for select is nil. I might
consider the 1.8.7 behavior a bug, but then again the underlying
select(2) syscall is allowed to have spurious wakeups. So even without
a timeout argument, select may always return even when nothing is
readable.

Thanks for the answer. My suspicion is that the Ruby 1.8 green threads
can be woken up, but the Ruby 1.9 system threads "do the right thing";
hence, the discrepancy in observed behavior. Good to know that it's
not a bug in 1.9.

/me goes off to rework code

Yeah, I expect thread.wakeup to only work on thread-aware things like
sleeping on a condition variable or Thread.sleep. 1.8 wraps select()
and uses that as the basis of its green thread implementation, so its
harder to get around spurious wakeups. For reliably waking up IO#select
I'd just write a byte to a pipe (useful with signal handlers, too, see:
http://cr.yp.to/docs/selfpipe.html)

 

[Robert Klemme]

Yeah, I expect thread.wakeup to only work on thread-aware things like
sleeping on a condition variable or Thread.sleep. =A01.8 wraps select()
and uses that as the basis of its green thread implementation, so its
harder to get around spurious wakeups. =A0For reliably waking up IO#selec= t
I'd just write a byte to a pipe (useful with signal handlers, too, see:
http://cr.yp.to/docs/selfpipe.html)

Another approach would be to use #select with a timeout and regularly
check some condition.

Kind regards

robert

--=20
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/

 

[Tim Pease]

=20
Another approach would be to use #select with a timeout and regularly
check some condition.
=20

Are you done yet? Are you done yet? Are you done yet? Are you done yet?

I thought about that, but it's a little more code than I want to write. =
And I'd like my programs to *not* act like my 3 year old when he's tired =
Closing the socket is my solution. This is shutdown code ... just =
need my thread back so I can do some cleanup.

Thanks for the input.

Blessings,
TwP=

 

[Tim Pease]

to work in ruby 1.9 Can someone please confirm that this is the case. =
Is this the intended behavior, or is this a bug?

=20
=20
cat a.rb
require 'socket'
=20
pair =3D Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t =3D Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join
=20
ruby --version
ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]
=20
ruby a.rb
Thread is about to exit ...
=20
ruby1.9 --version
ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]
=20
ruby1.9 a.rb #=3D> hangs forever!!!!
=20
=20
=20
Any thoughts out there ???
=20
Hi Tim,
=20
1.9 is actually doing what I expect it to do, that is waiting
indefinitely because the timeout argument for select is nil. I = might
consider the 1.8.7 behavior a bug, but then again the underlying
select(2) syscall is allowed to have spurious wakeups. So even = without
a timeout argument, select may always return even when nothing is
readable.

=20
Thanks for the answer. My suspicion is that the Ruby 1.8 green = threads
can be woken up, but the Ruby 1.9 system threads "do the right = thing";
hence, the discrepancy in observed behavior. Good to know that it's
not a bug in 1.9.
=20
/me goes off to rework code

=20
Yeah, I expect thread.wakeup to only work on thread-aware things like
sleeping on a condition variable or Thread.sleep. 1.8 wraps select()
and uses that as the basis of its green thread implementation, so its
harder to get around spurious wakeups. For reliably waking up = IO#select
I'd just write a byte to a pipe (useful with signal handlers, too, = see:
http://cr.yp.to/docs/selfpipe.html)
=20

Oooooohhhhh ... rainbows and unicorns, indeed!! Great link.

/me goes off to rework code (again)

Blessings,
TwP=

 

[Tim Pease]

to work in ruby 1.9 Can someone please confirm that this is the case. =
Is this the intended behavior, or is this a bug?

=20
=20
cat a.rb
require 'socket'
=20
pair =3D Socket.pair(Socket::AF_UNIX, Socket::SOCK_STREAM, 0)
t =3D Thread.new {
Kernel.select pair, nil, nil, nil
puts "Thread is about to exit ..."
}
t.wakeup
t.join
=20
ruby --version
ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin10]
=20
ruby a.rb
Thread is about to exit ...
=20
ruby1.9 --version
ruby 1.9.1p243 (2009-07-16 revision 24175) [i386-darwin10]
=20
ruby1.9 a.rb #=3D> hangs forever!!!!
=20
=20
=20
Any thoughts out there ???
=20
Hi Tim,
=20
1.9 is actually doing what I expect it to do, that is waiting
indefinitely because the timeout argument for select is nil. I = might
consider the 1.8.7 behavior a bug, but then again the underlying
select(2) syscall is allowed to have spurious wakeups. So even = without
a timeout argument, select may always return even when nothing is
readable.

=20
Thanks for the answer. My suspicion is that the Ruby 1.8 green = threads
can be woken up, but the Ruby 1.9 system threads "do the right = thing";
hence, the discrepancy in observed behavior. Good to know that it's
not a bug in 1.9.
=20
/me goes off to rework code

=20
Yeah, I expect thread.wakeup to only work on thread-aware things like
sleeping on a condition variable or Thread.sleep. 1.8 wraps select()
and uses that as the basis of its green thread implementation, so its
harder to get around spurious wakeups. For reliably waking up = IO#select
I'd just write a byte to a pipe (useful with signal handlers, too, = see:
http://cr.yp.to/docs/selfpipe.html)
=20

Based on the self-pipe page, here is the solution I came up with. =
Please, anyone, give comments and suggestions.

http://gist.github.com/241224

Blessings,
TwP

 

[Eric Wong]

Based on the self-pipe page, here is the solution I came up with.
Please, anyone, give comments and suggestions.

http://gist.github.com/241224

Hi Tim,

I think a length argument is is required for read_nonblock. And
you should probably only be reading from the receiver *after* the
select.

Also, the creation/check of Thread#[:select_signal_pipe] is
potentially racy. I would probably create it in the parent
like this:

pipe = IO.pipe
thr = Thread.new(pipe) do |pipe|
Thread.current[:select_signal_pipe] = pipe
pipe.last.syswrite('.') # wakeup parent once we know the parent can use it
# ...
end
pipe.first.readpartial 1 # blocks until it reads '.'
# after the readpartial, we know the following is set:
assert(thr[:select_signal_pipe] == pipe)
# ...

Unlike the self-pipe example, the basic rule of thumb to avoid driving
oneself nuts is that one set of threads/processes only does reads, and
another set only does writes. Reading and writing to one pipe from the
same thread/process anywhere outside of signal handlers gets really
confusing, really quickly :> And even with the self-pipe signal
handlers, all communication is one-way within any given control block.

I'd also be very defensive about spurious select() wakeups (and
Errno::EINTR, too, just in case). Pretty much any syscall that sleeps
like select() including (but not limited to) pthread_cond_wait, poll,
epoll_wait are all *potentially* susceptible to spurious wakeups, so you
could be exiting a thread when you don't intend to.