nvcc segfault in Code_Expansion phase

[graham_old]

Hi,

When attempting to compile the following minimal test case:

global void test(int n)

{

int a[n];

}

The nvcc compiler segfaults with the following message:

Signal: Segmentation fault in Code_Expansion phase.

(0): Error: Signal Segmentation fault in phase Code_Expansion – processing aborted

*** Internal stack backtrace:

/usr/local/cuda/open64/lib//be [0x6d1c6f]

/usr/local/cuda/open64/lib//be [0x6d28b9]

/usr/local/cuda/open64/lib//be [0x6d200d]

/usr/local/cuda/open64/lib//be [0x6d3256]

/lib/libc.so.6 [0x2b7bed394040]

/usr/local/cuda/open64/lib//be [0x5461f9]

/usr/local/cuda/open64/lib//be [0x5740fc]

/usr/local/cuda/open64/lib//be [0x570b3c]

/usr/local/cuda/open64/lib//be [0x571fb1]

/usr/local/cuda/open64/lib//be [0x570b28]

/usr/local/cuda/open64/lib//be [0x57573b]

/usr/local/cuda/open64/lib//be [0x576269]

/usr/local/cuda/open64/lib//be [0x57688c]

/usr/local/cuda/open64/lib//be [0x552722]

/usr/local/cuda/open64/lib//be [0x405443]

/usr/local/cuda/open64/lib//be [0x4061f1]

/usr/local/cuda/open64/lib//be [0x40752d]

/lib/libc.so.6(__libc_start_main+0xe6) [0x2b7bed37f5a6]

/usr/local/cuda/open64/lib//be [0x4038da]

nvopencc INTERNAL ERROR: /usr/local/cuda/open64/lib//be died due to signal 4

This is with the most recent version of the CUDA toolkit:

nvcc: NVIDIA ® Cuda compiler driver

Copyright © 2005-2009 NVIDIA Corporation

Built on Fri_Feb_19_19:12:59_PST_2010

Cuda compilation tools, release 3.0, V0.2.1221

I suspect that this is not actually allowed anyway, and automatic arrays must have bounds that are constant at compile time. Is this correct?