Bug #13609: CI error on travis at spec for Process.setpriority

• Status: Closed
• Assignee: nobu (Nobuyoshi Nakada)
• Target version: -
• ruby -v: ruby 2.5.0dev (2017-05-29 trunk 58953) [x86_64-cygwin]
• Backport: 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN

[duerst (Martin Dürst)]

Trying to make a commit (r58954) for issue #13590, I got a CI error at https://travis-ci.org/ruby/ruby/builds/237086017. Reverting at r58955, the problem persisted at https://travis-ci.org/ruby/ruby/builds/237090520.

The error looked unrelated to the commit, and that it is not gone when reverting shows that it is indeed unrelated, so I'm raising this issue. For details, please see below.

The command "make -s $JOBS test-all -o exts TESTOPTS='-q --color=never --job-status=normal'" exited with 0.

7.28s$ make -s $JOBS test-rubyspec MSPECOPT=-j

generating x86_64-linux-fake.rb

x86_64-linux-fake.rb updated

ruby 2.5.0dev (2017-05-29 trunk 58955) [x86_64-linux]

[- | ==================100%================== | 00:00:00]      0F      0E 

1)

Process.setpriority sets the scheduling priority for a specified process group FAILED

Expected 0

 to equal 1

/home/travis/build/ruby/ruby/spec/rubyspec/core/process/setpriority_spec.rb:29:in `block (3 levels) in <top (required)>'

https://travis-ci.org/ruby/ruby/builds/237090520


/home/travis/build/ruby/ruby/spec/rubyspec/core/process/setpriority_spec.rb:3:in `<top (required)>'

Finished in 7.106548 seconds

3569 files, 26410 examples, 201546 expectations, 1 failure, 0 errors, 0 tagged

make: *** [yes-test-spec] Error 1

The command "make -s $JOBS test-rubyspec MSPECOPT=-j" exited with 2.

Done. Your build exited with 1.

[nobu (Nobuyoshi Nakada)]

• Description updated (diff)

It occurs sporadically in these days.

It seems a race condition between getpriority and setpriority in a process group.

# bug-13609.rb
Process.setpgrp
pr = Process.getpriority(Process::PRIO_PGRP, 0)
fork {Process.getpriority(Process::PRIO_PGRP, 0)}
Process.setpriority(Process::PRIO_PGRP, 0, pr+1)
unless Process.getpriority(Process::PRIO_PGRP, 0) == pr+1
  raise ARGV[0]
end

$ ./miniruby -v; i=0; while [ $i -lt 100 ] && ./miniruby bug-13609.rb $[++i]; do :; done
ruby 2.5.0dev (2017-05-29 trunk 58958) [x86_64-linux]
Traceback (most recent call last):
bug-13609.rb:7:in `<main>': 24 (RuntimeError)

I'm not sure if this is an expected behavior, I think this test should be run in a separate process group.

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Closed

Applied in changeset trunk|r58959.

---

Run in separate process/pgrp [Bug #13609]

[Eregon (Benoit Daloze)]

Thanks for the fix nobu!