You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
LocalStore::addToStore() since
79ae9e4 expects a regular NAR hash,
rather than a NAR hash modulo self-references. FixesNixOS#6300.
Also, makeContentAddressed() now rewrites the entire closure (so 'nix
store make-content-addressable' no longer needs '-r'). See NixOS#6301.
Since the hash rewriting is only a heuristic, it seemed safer to provide a local escape hatch (saying that a specific derivation should be input-addressed while all the rest would be CA);
Enforcing everything to be CA can make it hard to interact with external derivations − for example my system config is globally content-addressed, but I also include some packages from a few external flakes (and which are input-addressed).
Now, it seems that 1. was essentially a theoretical concern (there has been a couple of CA-related breakages in nixpkgs, but all easy to fix and due to #4764 and not the hash rewriting) so that’s not a really strong point anymore.
For 2, maybe we could expose makeContentAddressed as a builtin so that we can rewrite on-the-fly any input-addressed derivation output that we’d want to use?
LocalStore::addToStore() since
79ae9e4 expects a regular NAR hash,
rather than a NAR hash modulo self-references. FixesNixOS#6300.
Also, makeContentAddressed() now rewrites the entire closure (so 'nix
store make-content-addressable' no longer needs '-r'). See NixOS#6301.
Currently CA paths can refer to non-CA paths:
This seems like an undesirable property, since it means that a top-level CA path does not "lock" its dependencies.
@thufschmitt What do you think?
The text was updated successfully, but these errors were encountered: