Louis Dionne

Looking at the diff between 2.0.9 and 2.0.10 [here](https://github.com/rtomayko/tilt/compare/v2.0.9...v2.0.10), I would be tempted to say that the culprit is https://github.com/rtomayko/tilt/commit/be625c89a9e3717fe23f90f9ded0c8a503b66a7d.

FWIW, I'm personally fine with the suggested workaround of using `def self.foo`.

> Finally, compile-times. If you are doing a lot of meta-programming Hana (and other libraries) compile faster than meta. For correctness: I don't think that's true. Hana is actually heavier...

I'm sorry, I don't have the bandwidth to provide a PR for fixing this at the root in meta. That's why I just put a band-aid by fixing metabench. Also,...

I understand that; I was rationalizing my initial closing of the issue.

You can prepare the pack of indices you want to take in O(1) (using `__make_integer_seq`, and then expand into a list using the `__type_pack_element` intrinsic explained [in a blog post](http://ldionne.com/2015/11/29/efficient-parameter-pack-indexing))....

Ahhh. Actually, I think this is a bug with the Boost distribution. When you install Boost, it should install a `lib/cmake/Hana/HanaConfig.cmake` file with it, but it doesn't. Reopening to keep...

> To make it work I used your script in the dependencies folder ! Make sure you're not using both that script AND a global installation of Boost, since that...

I acknowledge that this problem is a pain. You can also use `DYNO_STRING(foo)` instead of `"foo"_s`. Does that solve your problem, or are you set on using the UDLs but...

Would you suggest the `dyno::literals` namespace be dropped in that case? I've made the change of `_s` to `_dyno` on a branch locally and it's a trivial change for me...