fetchgit ignore deepClone and does a shallow clone anyway

[jbgi]

Since #130040 deepClone = true; (for fetchgit and fetchFromGitHub) and nix-prefetch-git --deepClone don't work anymore.
ie, when running

cat $(nix-prefetch-git --quiet --deepClone --leave-dotGit https://github.com/NixOS/rfcs 220af8e4f171425d168334f07163ea3db2eb8161 | jq -r .path)/.git/shallow

Actual result :

220af8e4f171425d168334f07163ea3db2eb8161

Expected:

cat: /nix/store/h1rfypna3gfbxnkhza30yjd792wr0id5-rfcs-220af8e/.git/shallow: No such file or directory

Reverting #130040 fix the issue, but the optimization should be kept when deepClone is not used.

[Atemu]

Could you give a real-world scenario for deepClone? All the ones in Nixpkgs are broken, with or without the optimisation.

AFAICT your reproducer merely checks whether a clone is sparse?

[jbgi]

My real-world scenario is for a short-lived, local only derivation where the broken aspect (non-determinism) of deepClone are ok.

My issue is that the doc say:

If deepClone is set to true, the entire repository is cloned as opposing to just creating a shallow clone.

but my reproducer shows that this is not the case: --deepClone produce a shallow clone. This should either be fixed, or --deepClone be removed entirely if we don't want to support it. I don't really care either way, I can use 21.05 for the non (completely) broken deepClone support, and, anyway, my use-case is mostly for some convenience during testing a "merge multiple repos" script, but nothing I can't live without.

[Atemu]

What I need is a concrete Nix example snippet that makes use of a property of deepClone and doesn't work without because I don't know what deepClone is useful for.
I need to be able to understand what the script did before my patch, why it's different now and how that affects real-world deepClone usage.

Relying on a stable version is not a good idea as that will be deprecated in half a year or so and will likely break in the future. My change should also not have caused any regressions in the real world.

Given that it's taken this long for someone to find out though, I also question why we even need deepClone. I also don't really understand why you'd need it in the first place. It'd be great if you, as a real-world consumer of that feature, could enlighten me of that

[jbgi]

Well the point of using deepClone is to retain the full git history, my use case is merging multiple repos into one, while preserving history, which is why I needed deepClone to work (though I could work around it by not doing this in a nix derivation). It's here :
https://github.com/input-output-hk/cardano-world/blob/prepare-monorepo/nix/cardano/prepare-mono-repo/default.nix

#130040 force use of --depth=1 (shallow clone), even if a deepClone is asked for.

[wcole3]

Bumping to see if anyone has workarounds for this. Some projects make use of the contents in the .git folder and without deep clones they cannot be built cleanly. An example of a repo that makes use of the .git folder is https://github.com/PX4/PX4-Autopilot

[emily-is-my-username]

@wcole3 Disabling the optimized shallow clone when --deepClone is explicitly requested worked for me (patch here)

[infinisil]

@emily-is-my-username I created a PR from your patch! I hope you don't mind: #252865