Backport #7159

SIGSEGV when calling Random#rand after forking

Added by Kristóf Marussy about 4 years ago. Updated about 4 years ago.



When Kernel#fork is called,
Ruby uninitializes the default random generator Random::DEFAULT
by setting the next field of the struct MT that provides its state to 0.

Kernel#rand correctly initializes this random generator by a seed
when it detects this.
However, Random#rand does not do this and therefore attempting to
call Random#rand in a newly forked child process results in a
segmentation fault, because a null pointer is being dereferenced.

This irb session illustrates the problem:

Related issues

Duplicates Backport193 - Backport #5661: Segfault in Random.rand with Spork gem Closed 11/23/2011


#1 Updated by Nobuyoshi Nakada about 4 years ago

  • Tracker changed from Bug to Backport
  • Project changed from Ruby trunk to Backport193
  • Status changed from Open to Assigned
  • Assignee set to Usaku NAKAMURA

#2 Updated by Usaku NAKAMURA about 4 years ago

  • Status changed from Assigned to Closed

This issue was solved with changeset r37208.
Andrew, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

merge revision(s) 34977: [Backport #5661]

* random.c (random_s_rand): ensure default PRNG is re-initialized
  after fork.  patched by Eric Wong.  [Bug #5661]

Also available in: Atom PDF