| 6 Oct 2025 |
 fzakaria | still skips stuff | 22:44:01 |
 Sergei Zimmerman (xokdvium) | ma27: thanks for the gdb fix btw. Very much appreciated to be able to use boost's pretty-printers out of the box. I hate how many hoops one has to jump through to get rid of -Wa,--compress-debug-sections though | 23:21:53 |
| 7 Oct 2025 |
 lovesegfault | could one reasonably write a nix evaluator without a GC and use refcounting, or would there be too many leaks due to cycles? | 04:07:54 |
 K900 | Definitely too many cycles without any form of GC | 04:16:48 |
| K900 | But you can probably get something out of refcounting compared to baseline | 04:17:27 |
| lovesegfault | that's a shame, i was thinking that conceptually it'd be nice to write an evaluator around rust futures | 04:17:31 |
| lovesegfault | but no gc there | 04:17:55 |
| K900 | That would be insanely inefficient tbh | 04:18:57 |
| K900 | Well, I guess it depends on how big your future state machine gets | 04:19:24 |
| K900 | But I'd expect it to get pretty big | 04:19:30 |
| lovesegfault | yeah idk how it play out, just seemed like something fun to try | 04:20:24 |
| lovesegfault | maybe if i used something like https://crates.io/crates/sdd | 04:24:06 |
| lovesegfault | oh, hadn't seen this one before nix: ../canon-path.cc:82: void nix::CanonPath::push(std::string_view): Assertion c != "." && c != ".."' failed.` | 04:27:49 |
| lovesegfault | cc. Sergei Zimmerman (xokdvium) | 04:27:53 |
| lovesegfault | * oh, hadn't seen this one before nix: ../canon-path.cc:82: void nix::CanonPath::push(std::string_view): Assertion \c != "." && c != ".."' failed.` | 04:51:16 |
| lovesegfault | * oh, hadn't seen this one before nix: ../canon-path.cc:82: void nix::CanonPath::push(std::string_view): Assertion `c != "." && c != ".."' failed. | 04:51:49 |
| lovesegfault | i think this fixes it: https://github.com/NixOS/nix/pull/14171 | 05:00:22 |
 dramforever | anyone happen to know what this is fixed in? seems fixed on master, broken on 2.31-maintanance, but idk what this is
nix-repl> builtins.getFlake "github:NixOS/nixpkgs?narHash="
nix: ../flake.cc:37: nix::StorePath nix::flake::copyInputToStore(nix::EvalState&, nix::fetchers::Input&, const nix::fetchers::Input&, nix::ref<nix::SourceAccessor>): Assertion `!originalInput.getNarHash() || storePath == originalInput.computeStorePath(*state.store)' failed.
| 05:52:47 |
| dramforever | * anyone happen to know what commit this is fixed in? seems fixed on master, broken on 2.31-maintanance, but idk what this is
nix-repl> builtins.getFlake "github:NixOS/nixpkgs?narHash="
nix: ../flake.cc:37: nix::StorePath nix::flake::copyInputToStore(nix::EvalState&, nix::fetchers::Input&, const nix::fetchers::Input&, nix::ref<nix::SourceAccessor>): Assertion `!originalInput.getNarHash() || storePath == originalInput.computeStorePath(*state.store)' failed.
| 05:53:00 |
| Sergei Zimmerman (xokdvium) | In reply to @k900:0upti.me
That would be insanely inefficient tbh Well, snix only uses Rc for Values iirc | 10:56:26 |
| K900 | Yeah and it needs like 200GB of memory to evaluate nixpkgs | 10:58:56 |
| Sergei Zimmerman (xokdvium) | Honestly if one is careful about the memory layout we could get parity with the current memory usage even without GC | 11:05:25 |
| Sergei Zimmerman (xokdvium) | In reply to @dramforever:matrix.org
anyone happen to know what commit this is fixed in? seems fixed on master, broken on 2.31-maintanance, but idk what this is
nix-repl> builtins.getFlake "github:NixOS/nixpkgs?narHash="
nix: ../flake.cc:37: nix::StorePath nix::flake::copyInputToStore(nix::EvalState&, nix::fetchers::Input&, const nix::fetchers::Input&, nix::ref<nix::SourceAccessor>): Assertion `!originalInput.getNarHash() || storePath == originalInput.computeStorePath(*state.store)' failed.
I think it’s #14050 | 11:06:38 |
| lovesegfault | so, last night i was playing with hooking up honggfuzz to libexpr | 14:39:31 |
| lovesegfault | and i think i got it working, and then went to sleep | 14:39:48 |
