Alexandre Amice

Can confirm that using Mosek to solve the program runs without crashing and returns that the program is infeasible.

FWIW, if one were to add this feature `iris.h` would then depend on `planning/collision_checker`. I specifically put the features of #20831 in planning to avoid having this dependency as having...

This issue is coming back up in my work with @Michaelszeng. The use case is being able to build IRIS regions with filtered collision geometries. As mentioned above, it would...

FWIW. I just got bit by this when writing the `IrisFromCliqueCover` code. In this code, I have IRIS working in many worker threads which works correctly in my C++ tests,...

> @AlexandreAmice Is there any chance you can share your code / diff / whatever here? It would be good to see a worked example so that we can figure...

cc @wernerpe @Michaelszeng

> Update: This error only seems to happen when Gurobi is used as a LP solver. I am not seeing any segfaults in Mosek or Clarabel, which is very interesting....

Yes I did not intend to say that the test does nothing. I meant more specifically that ConvexConfigurationSpace does not effectively check the mixing step option

I don't think 1. is a problem, I can replicate it in C++ fine on my machine. The bug happens because Mosek and Gurobi pick the affine hull basis to...

cc @hongkai-dai @RussTedrake @jwnimmer-tri for discussion on the design.