Bruno Levy

Found out ! (took me some time), The supports have tiny differences in coordinates, it is not a closed boundary that defines a volume. I could fix it by snapping...

Ooops, did not want to close the issue, needs to be tested before. expansion::assign_product() allocates subproducts, on stack if number of components are smaller than `MAX_CAPACITY_ON_STACK`, that was set to...

Seems to fix the problem (at least for last run). Since the problem is not reproducible, let us wait for a couple of days before closing the issue...

Launched the tests three times, and still got a crash... Will need finer analysis of the crash.

Lowered max expansion size to 256 - it seems that it no longer crashes (to be tested with multiple runs) - it assertion-fails (trying to allocate an expansion on stack...

CSG seems to be OK on mac now, but we got another test that crashed on mac in debug mode, `compute_RVD` with `tordu.mesh` (still assertion failure `capa it is because...

Nope, 512 is too much. Well in fact we need to make a difference between: - when we need to switch from stack to heap - in `expansion` functions (product...

- Back to 256 on Macs - Made assertion in bytes_on_stack() less stringent, because standard predicates can use expansions as long as 400-500 components (without any problem because there is...

All light greens.... Let us test several times before closing the issue.

Ran the test three times, seems to be OK, Let us wait for nightly builds before closing the issue...