QP solver
In the force distribution, a QP problem needs to be solved, which raises the question of which library to use.
So far, everybody has been copying code from one repo to the other. The are talks in TSID to externalize it, but so far nothing has been done.
I really do not want to copy-paste a QP solver routine inside sot-talos-balance. We should rely on an external library for that. This is my belief. And I would like not to depend on TSID.
I started with eigen-quadprog, taken from JRL's repo. It is available on robotpkg's apt repositories, that's how I installed it. It worked well on a simple test and it seems to be doing fine in the SoT, given a very simple program. However, many people have raised concerns on this repo. It is based on C++ code which was auto-generated from Fortran. @ostasse explicitly warned me against using it. The repo was made 4 years ago with 3 commits and has never been touched since then. It should probably not be trusted.
Another option is eigen-qld, also from JRL. This is not on robotpkg, so it has to be installed from source. I couldn't compile the Python bindings, so I disabled them, we don't need them. The interface is more or less the same as eigen-quadprog. I also made a test, it seems ok. eigen-qld repo is also based on auto-generated code. However, the repo seems to be much more active than eigen-quadprog. We can probably trust it more.
Caron is currently using eigen-lssol, but it is not an option for us, because it relies on LSSOL, which is proprietary and needs a licence (also, the eigen-lssol itself is not publicly available and we should ask it from them). He is planning to switch to eigen-qld.
Finally, another possibility is qpOASES, but I haven't taken a look at it.
Rebus sic stantibus, I think I'll switch to eigen-qld and stick to it. @ostasse what do you think?