gchristensen changed the topic of #nixos-borg to: https://www.patreon.com/ofborg https://monitoring.nix.ci/dashboard/db/ofborg?refresh=10s&orgId=1&from=now-1h&to=now "I get to skip reviewing the PHP code and just wait until it is rewritten in something sane, like POSIX shell. || https://logs.nix.samueldr.com/nixos-borg
orivej has quit [Ping timeout: 250 seconds]
jtojnar has joined #nixos-borg
orivej has joined #nixos-borg
orivej has quit [Ping timeout: 250 seconds]
orivej has joined #nixos-borg
orivej has quit [Ping timeout: 250 seconds]
orivej has joined #nixos-borg
<infinisil> What was the reason for hash mismatches on darwin only again? #57659
<{^_^}> https://github.com/NixOS/nixpkgs/pull/57659 (by feijoas, 3 weeks ago, open): flat-remix-icon-theme: init at 2019-03-14
orivej has quit [Ping timeout: 258 seconds]
orivej has joined #nixos-borg
<LnL> unpacked hashing + unicode normalisation
<gchristensen> we might be able to replicate that on NixOS I think with ZFS's unicode normalization options https://docs.oracle.com/cd/E23824_01/html/821-1448/gazss.html
<gchristensen> (search normalization)
<LnL> yeah that's possible
<gchristensen> should Nix be doing unicode normalization on its own?
<gchristensen> seems a bit weird, but maybe yes?
<LnL> it could, but it's currently basically the same as tar if I understand it correctly
<gchristensen> right
<LnL> in general I don't really like using fetchFromGitHub if possible, in this case eg. https://github.com/daniruiz/flat-remix/archive/20190109.tar.gz
<gchristensen> yikes -- because we've pushed a lot of uses over to fetchFromGitHub :)
<LnL> oh wait, that's not a static dist
<LnL> yeah, I'm not even sure why. It's more work to split up an url into arguments then adding /archive IMHO
<samueldr> archives from github, those generated, are not stable; they did and may change
<samueldr> fetchFromGitHub uses those, but hash the contents to work around the issue of the archive changing :/
<infinisil> Hmm so what's the solution to the unicode normalization problem? Just ignore darwin?
<samueldr> though I think you knew
<LnL> that the easiest, but usually it's also possible to make the hash stable with a postFetch
<gchristensen> what does that sort of postFetch look like?
<gchristensen> ah cool
<gchristensen> I wonder if we could write a tool to do unicode normalization?
<gchristensen> and enforce it in the fetcher
<LnL> there's already something weird with hashing for darwin in nix, but I'm not sure how what it is
<LnL> use-case-hack Whether to enable a Darwin-specific hack for dealing with file name collisions.
<gchristensen> right :/
<LnL> but it's probably not possible to do something similar here since information is lost with the normalisation
<gchristensen> it is?
<gchristensen> maybe I'm misunderstanding what unicode normalization does
<LnL> dunno :)
* gchristensen DDG searches horrible things like "unicode normalization bash"
<LnL> how I think it works you shouldn't get hash mismatches so I have no idae
<gchristensen> find, xargs, uconv, mv , ...
<LnL> hehe
<LnL> "APFS stores the filename as it is received. No transformation is applied"
<samueldr> IIUC, it is nix that handles hashing the extracted directory, right? could it be a feature added to nix to hash from a tarball stream, while normalizing what needs to be normalized?
<LnL> that's what I thought, so list gives the same value but read/write is normalized before hashing
{`-`} has joined #nixos-borg
orivej has quit [Ping timeout: 240 seconds]