Fix build of implementation only opamParserTypes in Makefile build

[sternenseemann]

I had some issues linking against opam-file-format when it was built using the GNU make build process. I'm not sure if my solution is the best, but it has a pretty small diff and seems to work (I tested linking with building opaline).

[Having the make build process in place is kind of important for NixOS, since we'd otherwise run into circular dependencies since everything processing .install files depends on this library and we need to install it somehow which the Makefile nicely solves.]

[dra27]

Thanks! This is an alternate fix to the one proposed in #35 - I think that renaming the .mli will be the best way, but not completely sure yet!

[dra27]

I’m intrigued by the circular dependency - doesn’t using dune install deal with that?

[sternenseemann]

Thanks! This is an alternate fix to the one proposed in #35 - I think that renaming the .mli will be the best way, but not completely sure yet!

I've tried that, but I couldn't figure how to get it to work with the current build system (i. e. without touching the source), so I went with this solution.

I’m intrigued by the circular dependency - doesn’t using dune install deal with that?

I think it would, but we are using opaline for installing packages build by dune (and installing dune 1.x) which causes the circular dependency. I think I'll look into dune install soon-ish, since it is kind of annoying and I suspect the usage of opaline is for historical reasons.

[dra27]

This fix is in fact orthogonal to #35 - it does bring the Makefile build system inline with the current behaviour for Dune, but only inasmuch as they are then both equivalently broken. Compiling OpamParserTypes as a full module should fix this properly for both builds systems (and should be released this week)