Backport #2557

Segmentation fault in Marshal.load in Rails' ActiveSupport in 1.8.7-p248 only

Added by Rhett Sutphin about 2 years ago. Updated 9 months ago.

[ruby-core:27402]
Status:Closed Start date:01/05/2010
Priority:Normal Due date:
Assignee:Shyouhei Urabe % Done:

100%
Category:-
Target version:-

Description

There is a consistent segmentation fault in a particular invocation of Marshal.load inside of ActiveSupport.  This segfault happens on 1.8.7-p248 and not on 1.8.7-p174.  Apologies, but I have not been able to come up with a simpler reproduction.  Here's how you can reproduce it from activesupport:

 $ git clone git://github.com/rails/rails.git
 [...]
 $ cd rails/activesupport

Create a file called @fault.rb@ (or whatever you would like) with the following contents:

 # fault.rb
 require File.expand_path("../lib/active_support/multibyte/unicode_database.rb", __FILE__)
 ActiveSupport::Multibyte::UnicodeDatabase.new.codepoints
 puts "No crash"

Then execute it:

 $ ruby -v
 ruby 1.8.7 (2009-12-24 patchlevel 248) [i686-darwin9.8.0]
 $ ruby fault.rb 
 /private/tmp/rails/activesupport/lib/active_support/multibyte/unicode_database.rb:37: [BUG] Segmentation fault
 ruby 1.8.7 (2009-12-24 patchlevel 248) [i686-darwin9.8.0]

 Abort trap

But in 1.8.7-p174:

 $ ruby -v
 ruby 1.8.7 (2009-06-12 patchlevel 174) [i686-darwin9.8.0]
 $ ruby fault.rb
 No crash

This may be related to #2175.  Unfortunately, I can't read Japanese so I'm not sure.  The backtrace from gdb for this crash does not match the one in #2175, but a colleague pointed out that including @GC.disable@ in @fault.rb@ prevents the crash.

Related issues

duplicated by Backport87 - Backport #2597: Crash in p248 while installing gem Closed 01/12/2010

Associated revisions

Revision 27408
Added by shyouhei almost 2 years ago

merge revision(s) 26251: * ruby.h (RB_GC_GUARD_PTR): workaround for gcc optimization. [ruby-core:27402]

History

#1
Updated by Yui NARUSE about 2 years ago

  • Status changed from Open to Assigned
  • Assignee set to Nobuyoshi Nakada

#2
Updated by Nobuyoshi Nakada about 2 years ago

  • Status changed from Assigned to Closed
fixed by r26251.

#3
Updated by Nobuyoshi Nakada about 2 years ago

  • Status changed from Closed to Assigned
  • Assignee changed from Nobuyoshi Nakada to Shyouhei Urabe

#4
Updated by Rhett Sutphin almost 2 years ago

There was a deficiency in my original bug report.  I failed to mention that I had only run this test on OS X 10.5.8.  I have re-run it with p249 and from today's head version of 1.8.7.  The code still segfaults on OS X 10.5.8, but it does not fail on Linux (specifically Ubuntu with kernel 2.6.28-18-server).  Perhaps it is OS X / darwin only.

#5
Updated by Rodney Woodruff almost 2 years ago

Hi,

thanks for the post. I appear to be having the exact same problem after upgrading to 1.8.7 p249. I am on OS X 10.5.8 and use MacPorts. Is there a fix for this yet?

Best,

-- Rodney

#6
Updated by Greg Moreno almost 2 years ago

I am confirming that this happens in OSX 10.5.8 with 1.8.7 p249.  Here's my setup (via rvm info)

system:
  uname:        "Darwin gokou.dev 9.8.0 Darwin Kernel Version 9.8.0: Wed Jul 15 16:55:01 PDT 2009; root:xnu-1228.15.4~1/RELEASE_I386 i386"
  shell:        "bash"
  version:      "3.2.17(1)-release"

ruby:
  interpreter:  "ruby"
  version:      "1.8.7"
  date:         "2010-01-10"
  platform:     "i686-darwin9.8.0"
  patchlevel:   "2010-01-10 patchlevel 249"
  full_version: "ruby 1.8.7 (2010-01-10 patchlevel 249) [i686-darwin9.8.0]"

#7
Updated by Greg Moreno almost 2 years ago

I don't encounter the bug in Ubuntu 9.10 with 1.8.7 p249.

system:
  uname:        "Linux gohan 2.6.31-20-generic #58-Ubuntu SMP Fri Mar 12 05:23:09 UTC 2010 i686 GNU/Linux"
  shell:        "bash"
  version:      "4.0.33(1)-release"

ruby:
  interpreter:  "ruby"
  version:      "1.8.7"
  date:         "2010-01-10"
  platform:     "i686-linux"
  patchlevel:   "2010-01-10 patchlevel 249"
  full_version: "ruby 1.8.7 (2010-01-10 patchlevel 249) [i686-linux]"

#8
Updated by Shyouhei Urabe almost 2 years ago

  • Status changed from Assigned to Closed
  • % Done changed from 0 to 100
This issue was solved with changeset r27408.
Rhett, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

Also available in: Atom PDF