[ruby-core:98158] [Ruby master Bug#16835] SIGCHLD + system new behavior on 2.6

Issue #16835 has been updated by maths22 (Jacob Burroughs).


jeremyevans0 (Jeremy Evans) wrote in #note-2:
> I agree that based on a review of the commit message, this change was unintended, and since the issue was not known previously, it was not documented.
> 
> However, I think the new behavior makes more sense. `system` creates a child process, so when the child process exits, it should trigger a `SIGCLD` trap.  That is how `sh` itself works (example assumes `echo` is a shell built-in, and uses CHLD instead of CLD as that is the signal name on the operating system I am using):
> 
> ```
> trap "echo child exit" CHLD
> $ echo 1
> 1
> $ /bin/echo 1
> 1
> child exit
> ```
> 
> Other than compatibility with Ruby 2.5 and prior, is there a good reason to follow the historical behavior?

Going by https://github.com/ruby/ruby/commit/6b87ac68dba3d8967d76233766a174c8a82813e3 , it seems like it was a deliberate decision in the past to *not* have the SIGCLD signal bubble up. I only found this because it caused a rather hard to track down bug in our code, I don't think we actually intended to rely on this behavior.

----------------------------------------
Bug #16835: SIGCHLD + system new behavior on 2.6
https://bugs.ruby-lang.org/issues/16835#change-85397

* Author: maths22 (Jacob Burroughs)
* Status: Open
* Priority: Normal
* ruby -v: ruby 2.6.5p114 (2019-10-01 revision 67812) [x86_64-darwin19]`
* Backport: 2.5: UNKNOWN, 2.6: UNKNOWN, 2.7: UNKNOWN
----------------------------------------
In rubies < 2.5, the `system` command did not trigger sigchld signal traps.  On ruby >= 2.6 it does.  This appears to have been introduced by https://github.com/ruby/ruby/commit/054a412d540e7ed2de63d68da753f585ea6616c3 .  To observe the change in behavior, run the following code on ruby 2.5 and 2.6:
``` ruby
Signal.trap("CLD")  { puts "Child died" }; system("true")
```
On ruby 2.5 it won't print anything.  On ruby 2.6 it will print "Child died".  I believe this is an unintended, undocumented change in behavior, but I may be wrong about that.





-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>