[CM] Re: cannot run cm script on Linux SBCL

Fernando Lopez-Lezcano nando@ccrma.Stanford.EDU
17 Aug 2005 11:14:10 -0700


On Wed, 2005-08-17 at 11:04, Fernando Lopez-Lezcano wrote:
> On Wed, 2005-08-17 at 08:25, Rick Taube wrote:
> > > FC4 -- I guess the answer is to keep trying -- here I am
> > > typing in a shell, with interspersed unprintable commentary:
> > 
> > I wonder if you are being bitten by a memory randomization issue in 
> > fc4, see below (there seems to be lots of postings about problems in 
> > 0.9.3)
> > 
> > ---
> > 	From: 	  jsnell@iki.fi
> > 	Subject: 	[Sbcl-devel] Re: problem building 0.9.3 under Fedora Core 4 
> > using CMUCL 19b
> 
> I'm building using the 0.9.0 binary image. 

(and with gcc4)

> > 	Date: 	August 17, 2005 10:14:00 AM CDT
> > 	To: 	  patrick.may@intamission.com
> > 	Cc: 	  sbcl-devel@lists.sourceforge.net
> > 
> > On Wed, Aug 17, 2005 at 12:09:45AM +0100, Patrick May wrote:
> > > Near as I can tell, I've got some bad craziness related to threading
> > > going on.  Can anyone lend me a clue?
> > 
> > Are you compiling with gcc 4? There are some problems with threading
> > and gcc4 (possibly solved in HEAD). Using gcc 3 instead might help.
> > 
> > > *** glibc detected *** sbcl: malloc(): memory corruption: 0x096ee058 
> > > ***
> > 
> > The FC4 kernel has some patches for memory map randomization, which
> > (unlike the earlier versions) is automatically turned on for all
> > applications. For full details see the "Memory randomization problems
> > coming" thread on sbcl-devel. For now, either of the following solutions
> > should work:
> > 
> >   Disable randomization globally "echo 0 > 
> > /proc/sys/kernel/randomize_va_space"
> >   Start sbcl with "setarch i386 -R sbcl"
> 
> Hmmm, interesting... I was running on the Planet CCRMA low latency
> kernel, probably does not have those patches. 

Seems like it does have that feature (running 2.6.12-0.21.rdt on fc4),
/proc/sys/kernel/randomize_va_space exists and is set to "1". Must be
some other issue then.

-- Fernando