J'ai maintenant mon propre dépôt #Opam. Pourquoi ? Comment ?
Tout est là:
https://zoggy.good-eris.net/posts/2026-03-16-own-opam-repo.html.fr
#ocaml
Maxence Guesdon : Avoir son propre dépôt Opam

Opam's Nix system dependency mechanism

https://ryan.freumh.org/opam-nix.html

#ocaml #opam #nix

Opam's Nix system dependency mechanism

Here's a special little error. I'm wondering if my dune build isn't using what's on disk, but the checked in git code, and I'm getting an error:

`[NOTE] Ignoring uncommitted changes in /workspaces/haxe-compiler-dev/vendor/haxe (`--working-dir' not specified or specified with no
argument).`

Only, thing is, I've tried adding `--working-dir` but `dune` doesn't recognise it, and isn't telling me who it is that needs the argument.

Neither Google nor Bing are telling me anything.

fuck

edit (wait, nope, my bad)

#Opam #Dune #OCaml #Haxe

Dear everyone, we just published an article on our #opam-repository archival work https://blog.robur.coop/articles/2025-03-26-opam-repository-archive.html

Happy reading #OCaml #opam #sustainability

Robur's blog - Pushing the opam-repository into a sustainable repository

The main opam-repository was only ever growing by collecting all releases of all packages. We worked hard on reducing the load for all clients by archiving packages.

I've had a machine with a default opam switch and a certain pin. I removed the pin and that triggered an update of the default switch to a new compiler version.

I have reservations about the idea to upgrade the compiler version without an explicit request from the user to do exactly that, but removing a pin really should do just that — remove a pin. The current behavior feels like a bug.

#ocaml #opam

opam exec -- emacs aka
eval $(opam env)) ?
#opam #Emacs
List of packages -- 4170 items. 12.7%. All unavailable or uninstallable (due to missing dependency). See more at:
https://discuss.ocaml.org/t/opam-repository-archival-phase-1-unavailable-packages/ #OCaml #opam #degrowth
Opam repository archival, Phase 1: unavailable packages

It is my pleasure to announce below the list of opam packages that will move to the opam-repository-archive on January 1st 2025. In total there are 4170 opam files scheduled for being moved within 561 unique packages. This decreases the size of the opam-repository by roughly 12.7%. This list contains all packages (a) marked as “available: false” (which may have various reasons: security issue, source unavailable, … - best to look into the “git log” for the specific package for the reason), and ...

OCaml

there is no standalone #MirageOS file system (and many of our unikernels don't need one). So, in the case of opam-mirror, we use the ocaml-git memory implementation: Git_mem.

Git_mem is different in that #Git objects are simply stored in a Hashtbl. There is no cooperation point when it comes to obtaining Git objects from this Hashtbl. So let's return to our original advice:

don't wrap code in Lwt if it doesn't do I/O.

And, of course, Git_mem doesn't do I/O. It does, however, require the process to be able to work with Lwt. In this case, Git_mem wraps the results in Lwt as late as possible (as explained above, so as not to slow down our processes unnecessarily). The choice inevitably means that the right-hand side can no longer offer cooperation points. And this is where our problem begins: composition.
ocaml-git wants to be a low-level library for irmin. By this fact, high-level commands such as a (patience) diff, git status, etc. are not implemented.

As a MirageOS project, ocaml-git is system agnostic. However, it provides a git-unix package which uses UNIX syscall and is able to introspect a usual Git repository in a filesystem. However, ocaml-git handles only Git objects and does not populate your filesystem as git does. For example, Git_unix.Sync.fetch does not give you files fetched from the repository but only synchronizes .git with that repository
#opam
https://blog.robur.coop/articles/lwt_pause.html
https://github.com/mirage/ocaml-git/tree/a36c90404b149ab85f429439af8785bb1dde1bee

Robur's blog - Cooperation and Lwt.pause

A disgression about Lwt and Miou

Bummer when the projector kills a demo. Nevertheless, a great talk by Brady on spatial ensemble and scene perception. #opam #Psynom14
If you think your experiment conditions put subjects through hell, image this. #opam #Psynom14 http://t.co/jxDCveleyW