tigerjython: init at 2.39

[rcmlz]

Description of changes

Adding new package TigerJython https://tigerjython.ch/

TigerJython offers everything you need to go from programming beginner to professional. You will find a wide variety of tutorials and can get started right away in programming environments specially developed for you.

TigerJython runs on the Java Virtual Machine (JVM), although we also offer the browser-based online solution WebTigerJython and WebTigerPython. The design and development of TigerJython follow the guidelines of simplicity, understandability, and—naturally—Python's battery included philosophy with an extensive set of libraries.

Things done

Taking the existing .tar.gz file from https://tigerjython.ch/en/products/download wrapping it to run the .jar file using the current stable JRE.

There are no dependencies introduced - just wrapping a jar for convenience.

• Fits CONTRIBUTING.md.

I run nix-build -A tigerjython and then /nix/store/...-tigerjython-2.39/bin/tigerjython successfully.

[Noodlez1232]

Result of nixpkgs-review pr 316431 run on x86_64-linux 1

1 package failed to build:

• tigerjython

[Noodlez1232]

The failure here was a hash mismatch. Probably the solution here is to update the hash in the source for TigerJython.tar and for the tjlogo64.png files.

[TomaSajt]

I realized that this problem is not that uncommon across nixpkgs,
and a good solution I found was using http://web.archive.org/ to cache a download link.

I did that here: http://web.archive.org/web/20240119124245/https://tigerjython.ch/user/pages/download/TigerJython.tar.gz

You'll have to generate another snapshot when there's a new release, but this is much easier to maintain than a custom git repo.

[rcmlz]

That is a good idea. However, I am not sure if that approach is a stable enough solution for a 150MB download file.

However, I took your advice and automated the creation of the archive using Github Actions and put the docs for that on the front page of https://github.com/zero-overhead/tj.

That way we get booth a.) simple way to create an archive, b.) high availability and bandwidth in case a whole class of e.g. 24 pupils is running a nix-shell -p tigerjython simultaneously or, even better, uses the home-manager config supplied by the teacher ;-)

What do you think? Is that a way forward?

PS: I recon I could even extend this GH action to update the package.nix, test it and send a pull request. But that would be an optimization I would only consider later, and for sources that release frequently new versions and were it is important to be up to date.

TigerJython is very mature (and I assume mostly used in schools in the DACH area anyway). Using an older version is not a security risk - pupils just might not be able to connect the latest Robot or Platform or use the new actor X on model Y.

[rcmlz]

What can I do to make this PR moving forward?

[emilazy]

web.archive.org is used extensively throughout the tree (112 files in pkgs/ matching it), and doesn’t carry as many of the supply chain security risks as using a binary from an unreviewed maintainer‐controlled source. The Internet Archive regularly serves files of far greater size than that through their collections, and you can mark this package as unfreeRedistributable (“Any such service or product may be shared and distributed freely, provided the TJ Group is identified as the original copyright holder.”) to get nixos.org to cache it, so I think that’s a far preferable solution.

[rcmlz]

Thank you for your review. I was not aware that using archive.org is the standard way of doing it, furthermore I assumed that each installation will trigger a download from that side.

I changed the source accordingly - and use the same approach for a PR for another software (I like to be) used in schools - filius.

[TomaSajt]

and you can mark this package as unfreeRedistributable (“Any such service or product may be shared and distributed freely, provided the TJ Group is identified as the original copyright holder.”) to get nixos.org to cache it, so I think that’s a far preferable solution.

AFAICT unfreeRedistributable packages are not built on hydra...
#83884

[emilazy]

Sorry, you’re right; I got my wires crossed. Won’t tarballs.nixos.org still cache the source derivation though? I admit I haven’t spent much time working with non‐free packages, so I could be totally wrong here. In any case I’m pretty sure the Internet Archive can take the load.

[rcmlz]

Thank you for your help. I understood that the current state of the PR is as good as it gets? Can I improve something else? Do I need to edit any other files?

[TomaSajt]

I've requested some changes.

Also, could you squash your commits into 2 commits?

• one for adding yourself as a maintainer
• and one for adding the package itself

https://github.com/NixOS/nixpkgs/blob/master/CONTRIBUTING.md#commit-conventions
https://github.com/NixOS/nixpkgs/blob/master/pkgs/README.md#commit-conventions

[rcmlz]

Thank you. I incorporated the change requests. Can you give me a hint how I can squash the 3 commits into 2?

[TomaSajt]

Thank you. I incorporated the change requests. Can you give me a hint how I can squash the 3 commits into 2?

You can run git reset HEAD~3 to undo the last 3 commits, but keep the changes. You can then create the two commits from there.
Follow the commit conventions I linked previously.

[rcmlz]

@TomaSajt @emilazy @Noodlez1232 @LeSuisse - Thank you all for your help - graciously appreciated. I learned a lot for my next contribution. Now I am eager to see how/if I can use Nix in school to ensure unified computer setup in a BYOD world.

[vigress8]

Per NixOS/rfcs#166, format the file with

nix run nixpkgs#nixfmt-rfc-style -- pkgs/by-name/ti/tigerjython/package.nix

[vigress8]

Since updating this seems somewhat cumbersome, I wonder if an updateScript could be written to handle it.

[rcmlz]

@vigress8 thank you for your review. I included the suggested changes (and fixed a small bug in the java wrapper).

With regards to the suggested update script as a side note: TigerJython is mature and not releasing frequently. I included the steps a human (me) or a script would need to do.

Do you assume scraping https://tigerjython.ch/en/products/download for version string and download link is stable enough to implement?

My current view is that the risk of failures in webscraping is higher than the effort of a yearly (before Autumn term) manual update.

[vigress8]

My current view is that the risk of failures in webscraping is higher than the effort of a yearly (before Autumn term) manual update.

Fair enough, it's not quite fit for automation. But perhaps a script meant to run manually, and the result checked by a human, would be useful.

[TomaSajt]

The PR looks pretty complete now, though I still have some small suggestions.

[TomaSajt]

Also I'm a bit concerned about the long download times.
If someone used nix-collect-garbage, the tarball would probably get cleared out, since it's a build-time only dependency. Then, if the package needed to be rebuilt, it would need to fetch the tarball again, which is painfully long.
Let's hope whoever uses this package uses the stable branches, where things don't get rebuilt so often.

It might be possible to hack together something where we symlink the tarball's files instead of copying them into the main package. This way they should not get cleared. What do you think, is it worth trying?

[rcmlz]

It might be possible to hack together something where we symlink the tarball's files instead of copying them into the main package. This way they should not get cleared. What do you think, is it worth trying?

I am not able to hack together such a thing. Would the generic solutions to this situation not be - as suggested by you - to cache unfreeRedistributable packages? This was the Issue you linked.

I personally plan to use only the stable branch (School - things must work in class, no experiments on infrastructure during terms)

Alternatively I ask upstream if I can declare the package with a free license such that caching happens automatically?

[TomaSajt]

I've though about the download speed a bit more and I think it's fine as it is for now.

If someone has an issue with long download times when rebuilding, we can ask them to just manually download the tar.gz file and do (tigerjython.overrideAttrs { src = ./TigerJython.tar.gz; })

[TomaSajt]

LGTM for the most part.
If some issue comes up in the future, we can fix it then.

[JohnRTitor]

Result of nixpkgs-review pr 316431 run on x86_64-linux 1

1 package built:

• tigerjython

[rcmlz]

@LeSuisse - what do you think? Is this PR ready? To start TigerJython currently I use

export NIXPKGS_ALLOW_UNFREE=1; nix run github:nixos/nixpkgs/pull/316431/head#tigerjython --extra-experimental-features "nix-command flakes" --impure

but it would be more convinient to have it as a standard package.