Bug#812852: Fwd: Re: Vigra auf ppc64el

[Daniel Stender]

-------- Forwarded Message --------
Subject: Re: Vigra auf ppc64el
Date: Fri, 4 Mar 2016 15:56:16 +0100
From: Ullrich Koethe <ullrich.koethe at iwr.uni-heidelberg.de>
To: Daniel Stender <stender at debian.org>
CC: Francesco Biscani <bluescarni at gmail.com>

Hallo Daniel,

unfortunately, we can't investigate this problem ourselves because we 
don't have access to a ppc system. Here are some ideas what you could try:

* Enable another set of threading primitives by passing

    -DWITH_BOOST_THREAD=1

to cmake. If the error goes away, the problem might be in the threading 
bits of Vigra or (less likely) in an outdated version of std::thread.

* Remove all occurencies of threading::memory_order* in 
multi_array_chunked.hxx (or equivalently, replace them with 
threading::memory_order_seq_cst). This will force all atomic accesses to 
use sequentially consistent ordering, the most strict possibility.

* compile + run with various sanitizers. Specifically, you should look 
into the address, undefined behaviour and thread sanitizer. The first 
two can be enabled with the gcc/clang command line flags:

  -fsanitize=undefined,address

The third one with:

  -fsanitize=thread

Alternatively or in addition

   valgrind --tool=helgrind ./tests/test_executable
   valginrd --tool=drd ./tests/test_executable

might be useful.

* It would be helpful to know if the problem is deterministic or not. Do 
multiple runs of the same test produce the same errors?

Best
Francesco and Ulli