>> install-script: add an install script to the new topkg package, > > This could work for packages that ship a static install file. But what > happens for packages that generate their install file during the build?
The script needs to be sufficiently complex to deal with all possible configurations. Which is what ordinary makefiles do in there install target. The downside of the install-script approach is that we would need to maintain such scripts for all packages in the transitive build-depends closure of opam that use topkg, for instance also for cmdliner. >> perl-installer: write a replacement for opam-installer, for >> instance in perl. > > This is my preference. Also my preference now. I am discussing this build dependency cycle off-list also with Daniel Bünzli. There I learned that there is already an OCaml library that can read these .install files: opam-file-format. The perl-installer will therefore probably be written in OCaml. I am also discussing with Daniel the chance that upstream solves the problem for us by changing opam-installer such that it does not depend on all the build-dependencies of opam. The outcome here is not clear yet, but assuming upstream decides to go this way, how long are we willing to wait? Or, the other way round, how long can we wait before updating cmdliner and all the other packages that now depend on topkg? I guess waiting for 2-3 month would be fine, but if the delay would be longer, we should pursue our own solution? Hendrik
