Bug #11411: Crash in rb_gc_mark() during Rails app boot with GC.stress=1

• Status: Closed
• Assignee: -
• Target version: -
• ruby -v: ruby 2.2.2p95 (2015-04-13 revision 50295) [x86_64-darwin14]
• Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN

[benweint (Ben Weintraub)]

When booting a newly-initialized Rails 4.2.1 app under Ruby 2.2.2 with GC.stress enabled, I get a repeatable crash. The same crash happens at least as far back as 2.1.4 (haven't tried any earlier versions).

Steps to reproduce:

1. gem install rails -v 4.2.1
2. rails new stress
3. cd stress
4. bundle
5. echo 'GC.stress = 1' >config/initializers/000_stress.rb
6. bundle exec rails s

Expected results:
The application should not crash.

Actual results:
The app crashes with a SIGABRT and the message seen below.

For convenience, I've attached a tarball of a Rails app built using the above steps that can reproduce this issue.

I'm running on Mac OS X 10.10.4, haven't tried with other OSes.

$ be rails s
=> Booting WEBrick
=> Rails 4.2.3 application starting in development on http://localhost:3000
=> Run `rails server -h` for more startup options
=> Ctrl-C to shutdown server

/Users/ben/.gem/ruby/2.2.2/gems/rack-1.6.4/lib/rack/mime.rb:55: [BUG] rb_gc_mark(): 0x007f94ebce3c18 is T_NONE

(snip)

Abort trap: 6

[nobu (Nobuyoshi Nakada)]

• Is duplicate of Bug #10933: [BUG] object allocation during garbage collection phase added

[nobu (Nobuyoshi Nakada)]

• Is duplicate of deleted (Bug #10933: [BUG] object allocation during garbage collection phase)

[wanabe (_ wanabe)]

• Related to Bug #11352: Sample code Garbage collection dies with an address is T_NONE added

[nobu (Nobuyoshi Nakada)]

• File bug-11411.log bug-11411.log added
• Description updated (diff)
• Status changed from Open to Closed

Seems same as #11352.

[nagachika (Tomoyuki Chikanaga)]

The fixes are backported into ruby_2_2 branch at r51521 (for #11352).