Leo White

I don't think `.cmi` files should be positional arguments. Positional arguments should be for inputs that are essentially consumed to produce the output. That is not true for `.cmi` files.

@dbuenzli What I'm trying to get at is that all OCaml's source files and intermediate files get used as positional argument exactly once -- when they are the main focus...

I agree that it would be better to use a new parameter name rather than overloading `-I`.

> It seems to me that the line you are trying to draw is quite a brittle one. To me it seems both an obvious difference and a fundamental one....

Ok, one last attempt to express the condition in a way that makes it clear: When you call a compiler the positional arguments should be the files that are getting...

> The role of each file can be clearly identified by its extension so flagging them is redundant from a cli interaction perspective. That is only the case if the...

> The same could be achieved for inheritance by calling find_class_type on any inherited class types during approx_class. I think this would fix the issue and probably break no programs...

What about using this jq: ``` jq 'walk(if type == "object" and has("file") then (.file |= match("([A-Za-z.]+)$").string) else . end)' ``` It should replace all `file` paths with just their...

What was the command? It's probably some marshalling issue. I don't think we use any magic numbers yet to stop us from trying to unmarshal the wrong files.

Yeah, that's probably a good idea. Might need a bit of thought for the best way to combine the results. I'm not sure just adding them together makes sense for...