Spring 4.2.0 crashing when running rake

[fschwahn]

This happened to me when running rails db:migrate.

I have never seen this before, and I have been using spring daily for years. Maybe it is connected to the changes in 4.2.0 (namely #708).

I'm running spring server as a separate process, so I have logs / stacktrace to share:

15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [4120] [server] accepted client
15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [4120] [server] running command rake
15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [4120] [server] shutting down
15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [4120] [application_manager:development] stopping
15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [4120] [application_manager:test] stopping
15:50:59 spring.1 | [2024-04-26 15:50:59 +0200] [12426] [application:development] forced exit
15:50:59 spring.1 | /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:63:in `send_io': Invalid argument - sendmsg(2) (Errno::EINVAL)
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:63:in `block in run'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:44:in `block in with_child'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:22:in `synchronize'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:41:in `with_child'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/application_manager.rb:62:in `run'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/server.rb:70:in `serve'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/server.rb:51:in `block in start_server'
15:50:59 spring.1 | 	from <internal:kernel>:187:in `loop'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/server.rb:51:in `start_server'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/server.rb:45:in `boot'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/server.rb:14:in `boot'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/client/server.rb:10:in `call'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/client/command.rb:7:in `call'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/client.rb:30:in `run'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/bin/spring:49:in `<top (required)>'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/binstub.rb:5:in `load'
15:50:59 spring.1 | 	from /Users/fabian/.rbenv/versions/3.3.1/lib/ruby/gems/3.3.0/gems/spring-4.2.0/lib/spring/binstub.rb:5:in `<top (required)>'
15:50:59 spring.1 | 	from <internal:/Users/fabian/.rbenv/versions/3.3.1/lib/ruby/3.3.0/rubygems/core_ext/kernel_require.rb>:136:in `require'
15:50:59 spring.1 | 	from <internal:/Users/fabian/.rbenv/versions/3.3.1/lib/ruby/3.3.0/rubygems/core_ext/kernel_require.rb>:136:in `require'
15:50:59 spring.1 | 	from bin/spring:12:in `block in <main>'
15:50:59 spring.1 | 	from <internal:kernel>:90:in `tap'
15:50:59 spring.1 | 	from bin/spring:9:in `<main>'

[fmichaut-diff]

Also running into this issue on MacOS.

This happens to me when the spring server is running, but a change gets detected by the watchers (eg: changing a config/ file).

The next client to connect to the server will trigger that error, killing the server process and any live pre-existing client and will make the new client hang indefinitely on preload_status = application.gets.

This is very frustrating when working on a git project, cause every change of branch is likely to trigger a watcher, which breaks the spring server and kills any open session (eg: a rails console being used, open a new terminal to run a rails db:migrate, rails g migration, or even another rails c will kill the running console, and the new process will hang indefinitely and won't even respond to CTRL-C)

I've been running into that issue for months now, it would be great if it could get fixed.

Relates to #396, #663, looks like this issue is very resilient