In reply to @vengmark2:matrix.org

1. Absolutely. There are far too many packages for any individual to keep track of all the non-Python dependencies and missing dependencies.
2. Generally, yes, but if that's too much work simply having the dependencies at all is better than not having them.
3. If you can, that would be great. Splitting your work into multiple PRs would be appreciated, though, since the difficulty of reviewing scales by something like O(N²) for N lines.

Questions about buildSystem:

"urllib3": [
"setuptools",
{
"buildSystem": "flit-core",
"until": "2.0.2"
},
{
"buildSystem": "hatchling",
"from": "2.0.2"
}
],

In reply to @vengmark2:matrix.org

1. Absolutely. There are far too many packages for any individual to keep track of all the non-Python dependencies and missing dependencies.
2. Generally, yes, but if that's too much work simply having the dependencies at all is better than not having them.
3. If you can, that would be great. Splitting your work into multiple PRs would be appreciated, though, since the difficulty of reviewing scales by something like O(N²) for N lines.

Questions about buildSystem:
Given the following:
"urllib3": [
"setuptools",
{
"buildSystem": "flit-core",
"until": "2.0.2"
},
{
"buildSystem": "hatchling",
"from": "2.0.2"
}
],

If I build 2.0.2, does it pull setuptools too?

In reply to @k900:0upti.me
It needs to be set with until and/or from to be conditional

Thanks.
Also, if I have

"my_package": [
{
"buildSystem": "flit-core",
"from": "2.0.1"
},
{
"buildSystem": "hatchling",
"from": "2.0.2"
}
],

If I build 2.0.2, does it pull flit-core?

      pathParts = filter ({ prefix, path }: "NETRC" == prefix) nixPath; # deadnix: skip
      netrc_file =
        if (pathParts != [ ])
        then (head pathParts).path
        else "";

In reply to @matthewcroughan:defenestrate.it
netrc_file is always "" because nixPath is always [] in pure-eval

Thanks for the feedback. The Nix flake variant indeed was missing in my PR. I added a few paragraphs how to add env vars to the nix-daemon (i think this is the only working way for flakes). Could you take a look again?

btw, thanks for your awesome talks on Nix for Docker (the main reason my company is willing to try Nix) and for Riscv.

In reply to @truh:matrix.org
Does it need to be that strict in fixed output derivations? Private git repos work too.

In reply to @malteneuss:matrix.org

Thanks for the feedback. The Nix flake variant indeed was missing in my PR. I added a few paragraphs how to add env vars to the nix-daemon (i think this is the only working way for flakes). Could you take a look again?

btw, thanks for your awesome talks on Nix for Docker (the main reason my company is willing to try Nix) and for Riscv.