JRuby

EXCEPTION_ACCESS_VIOLATION

Details

  • Type: Bug Bug
  • Status: Closed Closed
  • Priority: Blocker Blocker
  • Resolution: Fixed
  • Affects Version/s: JRuby 1.1.4
  • Fix Version/s: JRuby 1.1.5
  • Component/s: None
  • Labels:
    None
  • Environment:
    Windows XP Pro SP3, Java 1.6.0_07-b06, HotSpot build 10.0-b23, mixed mode, sharing), AMD X2 5600, 2GB RAM
  • Number of attachments :
    1

Description

This code causes an EXCEPTION_ACCESS_VIOLATION

c = 1

while (true)
flag = 0
(1..20).each do |e|
flag += (c % e)
end
break if (flag == 0)
c += 1
end
c

Below is the message:

#

  1. An unexpected error has been detected by Java Runtime Environment:
    #
  2. EXCEPTION_ACCESS_VIOLATION (0xc0000005) at pc=0x6d8a0242, pid=2912, tid=2772
    #
  3. Java VM: Java HotSpot(TM) Client VM (1.6.0_02-b06 mixed mode, sharing)
  4. Problematic frame:
  5. V [jvm.dll+0x30242]
    #
  6. An error report file with more information is saved as hs_err_pid2912.log
    #
  7. If you would like to submit a bug report, please visit:
  8. http://java.sun.com/webapps/bugreport/crash.jsp
    #

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
William Gannon added a comment -

I forgot to mention, this is with build 7858

Show
William Gannon added a comment - I forgot to mention, this is with build 7858
Hide
Permalink
Charles Oliver Nutter added a comment -

I was able to reproduce this with SoyLatte JDK 6 (1.6.0_03-p3) on OS X. The problem seems to be in the "client" JIT compiler in the JVM, which we've had some trouble with before. Segfaults in the JVM are always the JVM's fault, but there may be something we can do to work around it.

In the short term, you can use the "server" VM by passing --server (or -server to the "java" command) which does not crash. I will file a bug with the Hotspot devs.

Show
Charles Oliver Nutter added a comment - I was able to reproduce this with SoyLatte JDK 6 (1.6.0_03-p3) on OS X. The problem seems to be in the "client" JIT compiler in the JVM, which we've had some trouble with before. Segfaults in the JVM are always the JVM's fault, but there may be something we can do to work around it. In the short term, you can use the "server" VM by passing --server (or -server to the "java" command) which does not crash. I will file a bug with the Hotspot devs.
Hide
Permalink
Charles Oliver Nutter added a comment -

Managed to reduce it to this:

while (true)
  100.times {|e|} 
end

I talked with the Hotspot guys and they confirmed it's the same cause as an issue we reported last spring. That issue is fixed in all the various codebases, but I don't believe those fixes have propagated to any released version of Hotspot yet. The bug ID is 6758445, though I couldn't find it in the public bug listings.

I'm looking into the generated bytecode and will hopefully have it repaired soon.

Show
Charles Oliver Nutter added a comment - Managed to reduce it to this: 
while (true)
  100.times {|e|} 
end
I talked with the Hotspot guys and they confirmed it's the same cause as an issue we reported last spring. That issue is fixed in all the various codebases, but I don't believe those fixes have propagated to any released version of Hotspot yet. The bug ID is 6758445, though I couldn't find it in the public bug listings. I'm looking into the generated bytecode and will hopefully have it repaired soon.
Hide
Permalink
Charles Oliver Nutter added a comment -

I have committed a fix in r7870 that resolves this and cleans up/improves the "while" logic that was the source of the problem. I'm going to add a test that exhibits the crash before marking this resolved.

Show
Charles Oliver Nutter added a comment - I have committed a fix in r7870 that resolves this and cleans up/improves the "while" logic that was the source of the problem. I'm going to add a test that exhibits the crash before marking this resolved.
Hide
Permalink
Charles Oliver Nutter added a comment -

Hmm, I could not reduce this to a case that wouldn't slow down the build substantially. Short to medium runs of a fixnum-delimited loop do not appear to crash, and the true version would obviously run forever. So I think we'll have to go without a test for this one. I will, however, ping the ASM guys about possibly adding this as a verification. That would help reduce the likelihood of it coming up in the future.

Show
Charles Oliver Nutter added a comment - Hmm, I could not reduce this to a case that wouldn't slow down the build substantially. Short to medium runs of a fixnum-delimited loop do not appear to crash, and the true version would obviously run forever. So I think we'll have to go without a test for this one. I will, however, ping the ASM guys about possibly adding this as a verification. That would help reduce the likelihood of it coming up in the future.

People

Vote (0)
Watch (1)

Dates

  • Created:
    Updated:
    Resolved:
Atlassian JIRA (v4.4.3#663-r165197) | Report a problem
Powered by a free Atlassian JIRA open source license for Codehaus. Try JIRA - bug tracking software for your team.