reproducing SBCL bug LP #1009267

Hello everyone,

I recently reported SBCL bug #1009267, entitled "(gc :full t) has no
effect under certain circumstances". The url is
https://bugs.launchpad.net/sbcl/+bug/1009267

The symptoms of this bug were that GC will fail very quickly under certain 
conditions, when large arrays (each amounting to a significant fraction of 
the available heap space), are allocated sequentially. Perhaps more 
significantly, running (gc :full t) does not succeed in reclaiming all the 
allocated memory after a certain number of arrays had been allocated. This 
reclaim failure does not require heap exhaustion.

Since encountering this issue on my home desktop machine, which runs 32 
bit Debian squeeze, I have been able to reproduce it on my laptop, which 
also runs 32 bit Debian squeeze. Additionally, a friend was able to 
reproduce it on 32 bit Ubuntu 12.04 with very similar symptoms. He sent me 
the transcript, but is not a CL user, so did not investigate extensively. 
I was unable to reproduce this on an AMD64 server. also running Debian 
squeeze, which I also have access to. I also got a report from a user on 
#lisp that he was unable to reproduce it on a 32 bit Ubuntu 11.10 system. 
However, he did not provide me with a transcript of this.

The test code went through some changes; the most recent version is 
included below.

I'm writing this to request that other people test this, either on 32 bit 
or 64 bit Linux. Both negative and positive confirmations would be of 
interest to me. I want transcripts of the session. Either post them to the