I mentioned I was interested to look at spin in the last scikit-learn developer meeting. One feed-back I got was the worry that it adds another layer of tooling. There is some concern to end-up in a situation where there are too many layers. For example, spin that calls make that calls a shell script, that calls a Python script, etc ...

I looked a bit to see whether there was some built-in spin functionality that would somehow give some visibility into which commands gets run (kind of make --dry-run for example). This would allow to understand a bit more what gets run behind the scenes and potentially do it without spin if needed e.g. to put together a reproducer.

For now what I found is that you can add a verbose flag by command and it is the responsability of the command implementation to put whatever makes sense behind this verbose flag (or even without this verbose flag by the way). The meson commands seems to be quite verbose by default (which is a good thing in this case) and show which commands gets run. If we implement custom commands we will have to do it.

Did I get this roughly correctly?