[Bug 1071209] Re: memtest86+ test #7 false positives (random number sequence error)

[solik]

Downloaded Ubuntu 12.10 i386 ISO, converted it to OS X dmg and "burned" it to the USB stick.
Boot from this USB stick.
Begin memory test.
In every DIMM combination and on various chipsets test #7 fails around 129M.
Chipsets was Intel P45 and Intel Q35, memory was DDR2 400, DDR2 333 running on 800 and 666 MHz.

Downloaded memtest86+ 4.20 from website – no fail.

Basically, latest stable Ubuntu ISO distributing with buggy memory test
utility.

-- 
You received this bug notification because you are a member of Ubuntu
Foundations Bugs, which is subscribed to memtest86+ in Ubuntu.
https://bugs.launchpad.net/bugs/1071209

Title:
  memtest86+ test #7 false positives (random number sequence error)

Status in Memtest86+:
  Unknown
Status in Release Notes for Ubuntu:
  Incomplete
Status in “memtest86+” package in Ubuntu:
  Fix Released
Status in “memtest86+” source package in Quantal:
  Fix Released
Status in “memtest86+” source package in Raring:
  Fix Released
Status in “memtest86+” package in openSUSE:
  Fix Released

Bug description:
  Trying to check the memory at newly bought notebook I found a bug in
  the memtest86+ version 4.20 in ubuntu 12.10, in ubuntu 12.04 it is ok.
  The bug is reported at least in Fedora and Opensuse. It is assumed
  that the bug is caused by the gcc-4.7.

  It is easily  reproducible - select the test #7 in memtest or just
  wait till it - starting from the 129Mb it will report a lot of errors.
  I have checked it on three different systems , two of them I use on a
  daily basis and would note if RAM is really bad.

  https://bugzilla.redhat.com/show_bug.cgi?id=805813

  http://lists.opensuse.org/opensuse-bugs/2012-09/msg04386.html

  --

  SRU Justification

  [Impact]
  Users of memtest86+ will get false positives of memory failures. This will cause users to suspect hardware and require unnecessary testing/headaches.

  [Test Case]
  Boot Ubuntu in Quantal/Raring. At GRUB select memtest86+. Wait until the 7th test. It will fail at the 7th test.

  [Regression Potential]
  The fix is just adding another register to clobber in the inline assembly routine. Because  this affects newer GCC versions, older releases aren't affected. Thus, if there are compiler changes we should re-test.

To manage notifications about this bug go to:
https://bugs.launchpad.net/memtest86+/+bug/1071209/+subscriptions