CGAL users discussion list

[Michael Vennettilli <>]

Hello,

I am finding a strange issue. I construct a pair of linear cell 
complexes, chull_lcc and shell. To compute chull_lcc, I randomly sample 
points on the sphere and compute the convex hull. For each triangle in 
chull_lcc, I create a vertex in shell at the triangle's barycenter. This 
proceeds for some number of iterations.

Basically, I am finding that the ordering of points within a linear cell 
complex depends on the compilation details (debug vs release), inside 
which function things are cleared (even if an empty object is cleared), 
and whether or not my code is run through Valgrind. All of this is done 
with a fixed random seed, and the set of points are the same regardless 
of the details, just their ordering is different. This makes me worry 
about potential memory issues later on. Is this expected behavior? Are 
there no guarantees about the ordering of points even when the random 
seed is fixed?

I have analyzed a lot of cases, but I think that going into detail here 
would make the email quite lengthy. Instead, I will provide a link to a 
repo containing a minimal example capable of reproducing the issue: 
https://github.com/MichaelVennettilli/Minimal_example_shell_error. The 
description explains how to run the script, how to interpret the 
provided data files, and environmental details.

I appreciate any assistance with this issue.

Thank you,

Mike