ETS 'set' table bug ... or not?

[Tomas Abrahamsson]

> Hi there.  While running some tests on a borrowed Pentium 4 Xeon
> RedHat 8.0 Linux box, I encountered a segmentation violation problem
> that puzzled me.  I'm not certain if it's a bug in ETS or if it is a
> problem with the hardware.  The hardware claims to have ECC memory, so
> it's unlikely (but not impossible, I suppose) that bad RAM is causing
> the problem.
>
> The code for test.erl is included below.  If I run the command line:
>
>     erl -noinput -s test do set insert_rand 7000 delete_all \
> 	traverse_forward -s erlang halt
>
> ... no problem.  If I use a bigger number than 7000, no problem
> ... until I use this number: 53999154.  Then I get a segmentation
> violation every time, apparently while the 'delete_all' part of the
> test is running.

It crashes with a segmentation fault, apparently during
delete_all, on our Sun Blade-1000 as well. It produces
a core file 2147483647 (== 2^31-1) bytes long. However,
here it doesn't crash for 53999154, but for a number
somewhere between 52750000 and 52875000, I didn't have
time to pinpoint it further.

(erts-5.2.3.7, Solaris 8, the machine has 4GB RAM.)


/Tomas