ack, i'll drop D8621 then.

create a new exception class instead of IndexError, it's unclear that the rest of the function is not actually meant to raise that exception.

I can tell you it's not supposed to raise outside of this function.
Otherwise, i would have documented it in the main docstring of the function.

Rebase

Adapt according to suggestion

create a new exception class instead of IndexError, it's unclear that the rest of the function is not actually meant to raise that exception.

So we must do something about the fetcher which does crazy things.
Either filter out those patch or simply integrate those without checking them... (i don't like much the latter).

cat /var/tmp/sources-unstable-full.json | jq . | grep -C5 https://github.com/hunspell/hunspell/commit/ac938e2ecb48ab4dd21298126c7921689d60571b.patch

  "outputHash": "0bwfksz87iy7ikx3fb54zd5ww169qfm9kl076hsch3cs8p30s8az",
  "outputHashAlgo": "sha256",
  "outputHashMode": "flat",
  "type": "url",
  "urls": [
    "https://github.com/hunspell/hunspell/commit/ac938e2ecb48ab4dd21298126c7921689d60571b.patch"
  ],
  "integrity": "sha256-XyENxkWaDcg0NAfQmarDyQTOS/ukLDf6jMfHg76eji8=",
  "inferredFetcher": "fetchpatch"
},
{

As this one is a "flat" output hash mode, it's actually a real mismatch computation.
This renders nothing to try and use nix-store --dump as [1] does.

Improve intent implementation

by "not actionable", I means that being False tells the loaders the checksum cannot be
checked by hashing the tarball, but does not tell them how to check them.

is_integrity_about_artifact=False (which should be called are_checksums_about_artifact btw)

Another data point about ^. It's not important for the guix manifest [1]. We can keep a
compatible behavior for it and slightly improve the listing behavior for nixpkgs as it's
important for those [2].

Adapt according to suggestion

With the gazillion of new diffs on top of the origin lister code, we can now also list
the nixpkgs-unstable-full.json manifests [1]

Drop no forgotten commit that could have been dropped early on.

Related to T3781
Related to T4538

what's been said before (with the right local commit inside this time...)

Deal with ftp targetting a file without extension which will trigger a head request and
fail. This will skip it as nothing simple can be done to detect the nature of the file
then.

Adapt according to review:

• let ftp urls through
• format log or exception message with <%s>

Should we try asking Nixpkgs devs to fix their manifest instead?

Adapt according to review.
That makes sense plus i forgot i already had that import so fine!

Another one bites the dust [1]

Wrong stuff again...:

and you don't need grouper() for that

Actually randomize and do it at the manifest reading step.

I forgot the main gist of it

For the content loader i have mostly checksums mismatches [1].
It seems the integrity from the manifest is either wrong (or some in-place update occurred in the respective servers [2])