Timeout::ExitException can be raised by nested timeouts

[trogd0r]

Behavior

Nested timeout blocks can result in a Timeout::ExitException being raised, rather than a Timeout::Error

How to reproduce

The below will sometimes raise a Timeout::ExitException, sometimes a Timeout::Error (roughly 50/50).

begin
 Timeout.timeout(2) do
 Timeout.timeout(2) do
 sleep 3
 end
 end
rescue Exception => e
 puts "raised a #{e.class}"
end

Cause (maybe)

Hacking up the Timeout module as follows:

module Timeout
 class Error
 def self.handle_timeout(message) # :nodoc:
 exc = ExitException.new(message)

 begin
 puts "yield #{exc.object_id} (#{message})"
 yield exc
 rescue ExitException => e
 puts "test #{e.object_id} vs #{exc.object_id} => #{exc.equal?(e)} (#{message})"
 raise new(message) if exc.equal?(e)
 puts "re-raise #{e.object_id}('#{e}') (#{message})"
 raise
 end
 end
 end
end

Timeout.timeout(2, nil, "outer timeout") do
 Timeout.timeout(2, nil, "inner timeout") do
 sleep 3
 end
end

shows that, when an ExitException is raised, the "inner" exception is being caught by the "outer" handle_timeout block. Suspicious this is the result of indeterminate Thread#raise behavior?

Environment

Ruby 3.3.5
Timeout 0.4.1
Ubuntu 24.04.1 running on WSL2

[eregon]

It sounds somewhat related to https://bugs.ruby-lang.org/issues/20314
I'll try to think about a fix when I have some time.

[hsbt]

FYI: ruby/ruby#10851

[sergio-bobillier]

👋 We are experiencing this issue in our setup. But our experiments show that this only happens if both timeout calls have the same timeout value, for example the following test:

RSpec.describe 'Timeout' do
 it 'times out' do
 expect do
 Timeout.timeout(2) do
 Timeout.timeout(2) do
 sleep 19
 end
 end
 end.to raise_error(Timeout::Error)
 end
end

Fails all the time with a Timeout::ExitException that RSpec is not able to rescue:

Failures:

 1) Timeout times out
 Failure/Error: output.__send__(method, *args, &block)

 Timeout::ExitException:
 execution expired
 # Backtrace omitted ...
 #
 # Showing full backtrace because every line was filtered out.
 # See docs for RSpec::Configuration#backtrace_exclusion_patterns and
 # RSpec::Configuration#backtrace_inclusion_patterns for more information.

Finished in 2.01 seconds (files took 0.13324 seconds to load)
2 examples, 1 failure

However, doing this change:

 Timeout.timeout(2) do
- Timeout.timeout(2) do
+ Timeout.timeout(2.1) do
 sleep 19

Makes the test pass 100% of the time:

.

Finished in 2.01 seconds (files took 0.1275 seconds to load)
1 example, 0 failures

[eregon]

Taking a look at this and with:

require 'timeout'

begin
 Timeout.timeout(2) do
 Timeout.timeout(2) do
 sleep 3
 end
 end
rescue Exception => e
 p e
end

I get this consistently:

$ ruby -v -Ilib repro_nesting1.rb
ruby 3.4.7 (2025-10-08 revision 7a5688e2a2) +PRISM [x86_64-linux]
#<Timeout::Error: execution expired>
repro_nesting1.rb:10:in 'Kernel#p': execution expired (Timeout::ExitException)
	from repro_nesting1.rb:10:in '<main>'

FWIW on 3.3 it's slightly different:

$ ruby -v -Ilib repro_nesting1.rb
ruby 3.3.5 (2024-09-03 revision ef084cc8f4) [x86_64-linux]
#<Timeout::Error: execution expired>
repro_nesting1.rb:10:in `p': execution expired (Timeout::ExitException)
	from repro_nesting1.rb:10:in `rescue in <main>'
	from repro_nesting1.rb:3:in `<main>'

So the bad thing is the exception happens on line 10 which is p e and is after the Timeout.timeout blocks.
This is due to having multiple Thread#raise, as explained in https://bugs.ruby-lang.org/issues/20314

[eregon]

I think one idea would be to avoid calling Thread#raise in between another Thread#raise by Timeout and the Request#finished call, maybe using a lock per Fiber or so.
I need to try it.

[eregon]

I made a draft PR to fix this: #91
It works correctly for this example:

require 'timeout'

begin
 Timeout.timeout(2) do
 Timeout.timeout(2) do
 sleep 3
 end
 end
rescue Exception => e
 p :HERE
 puts ">"+e.full_message
end
puts

begin
 Timeout.timeout(0.1) do
 p sleep 3
 end
rescue Exception => e
 p e.class
end

$ ruby -v -Ilib repro_nesting1.rb
ruby 3.4.7 (2025-10-08 revision 7a5688e2a2) +PRISM [x86_64-linux]
:HERE
>/home/eregon/code/timeout/lib/timeout.rb:40:in 'Timeout::Error.handle_timeout': execution expired (Timeout::Error)
	from /home/eregon/code/timeout/lib/timeout.rb:363:in 'Timeout.timeout'
	from repro_nesting1.rb:4:in '<main>'
repro_nesting1.rb:6:in 'Kernel#sleep': execution expired (Timeout::ExitException)
	from repro_nesting1.rb:6:in 'block (2 levels) in <main>'
	from /home/eregon/code/timeout/lib/timeout.rb:353:in 'block (2 levels) in Timeout.timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:161:in 'Timeout::FiberState#interruptible'
	from /home/eregon/code/timeout/lib/timeout.rb:209:in 'Timeout::Request#interruptible'
	from /home/eregon/code/timeout/lib/timeout.rb:352:in 'block in Timeout.timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:38:in 'Timeout::Error.handle_timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:363:in 'Timeout.timeout'
	from repro_nesting1.rb:5:in 'block in <main>'
	from /home/eregon/code/timeout/lib/timeout.rb:353:in 'block (2 levels) in Timeout.timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:161:in 'Timeout::FiberState#interruptible'
	from /home/eregon/code/timeout/lib/timeout.rb:209:in 'Timeout::Request#interruptible'
	from /home/eregon/code/timeout/lib/timeout.rb:352:in 'block in Timeout.timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:38:in 'Timeout::Error.handle_timeout'
	from /home/eregon/code/timeout/lib/timeout.rb:363:in 'Timeout.timeout'
	from repro_nesting1.rb:4:in '<main>'

Timeout::Error