| 15 Oct 2025 |
 fzakaria | * Like can it say "Git cache is out of date" or "Tarball expires" or w/e is causing the fetching source.
(I'm thinking it's a flake input fetch although it writes "source fetch") | 13:18:49 |
 Robert Hensing (roberth) | Sergei Zimmerman (xokdvium): wdyt? https://github.com/libgit2/libgit2/compare/main...roberth:libgit2:refs-shorthand-multi-segment?expand=1 | 16:00:58 |
 Sergei Zimmerman (xokdvium) | In reply to @roberthensing:matrix.org
Sergei Zimmerman (xokdvium): wdyt? https://github.com/libgit2/libgit2/compare/main...roberth:libgit2:refs-shorthand-multi-segment?expand=1 Hmm, I was thinking that maybe for our use-case we could also just use is_valid_branch_name in conjunction with is_valid_reference_name instead of the whole normalization dance? Though yeah, this does seem like an oversight in libgit2 as well | 16:33:40 |
| Robert Hensing (roberth) | I guess I could open that PR as an alternative if those two functions don't work out? | 16:35:03 |
| Sergei Zimmerman (xokdvium) | In reply to @roberthensing:matrix.org
I guess I could open that PR as an alternative if those two functions don't work out? Let’s see how upstream reacts. They’be been ignoring my DEL patch for quite some time now | 16:37:13 |
| Sergei Zimmerman (xokdvium) | In reply to @roberthensing:matrix.org
I guess I could open that PR as an alternative if those two functions don't work out? * Let’s see how upstream reacts. They’ve been ignoring my DEL patch for quite some time now | 16:37:29 |
| fzakaria | I'm on a machine with 120 cores and i'm surprised how low parallelism meson compile -j 100 is on it | 17:50:39 |
| fzakaria | * I'm on a machine with 120 cores and i'm surprised how low parallelism meson compile -j 100 is on it for compiling Nix | 17:50:46 |
| Sergei Zimmerman (xokdvium) | In reply to @fzakaria:one.ems.host
I'm on a machine with 120 cores and i'm surprised how low parallelism meson compile -j 100 is on it for compiling Nix PCH make for a pretty bad fork-join | 17:52:38 |
| Sergei Zimmerman (xokdvium) | Also the linker | 17:52:56 |
| fzakaria | Also, my clangd fails to find the boost header only dependencies? Is there a setting someone can recommend here | 17:52:57 |
| fzakaria | I can see they are missing from meson compile as they are not "true" dependencies.. | 17:53:12 |
| Sergei Zimmerman (xokdvium) | In reply to @xokdvium:matrix.org
Also the linker You should disable b_lto | 17:53:23 |
 Philip Taron (UTC-8) | At $WORK we use clang modules to escape the PCH trap | 17:53:25 |
| Sergei Zimmerman (xokdvium) | Honestly PCH is pretty good for us. LTO on the other hand is very slow | 17:54:04 |
| fzakaria | I think it's cause NIX_CFLAGS is implicit outside meson? | 17:54:06 |
| Sergei Zimmerman (xokdvium) | In reply to @philiptaron:matrix.org
At $WORK we use clang modules to escape the PCH trap Let’s just wait until 2036 for modules to get supported on all platforms and in meson | 17:55:41 |
| Philip Taron (UTC-8) | Our impl doesn't give two hoots about gcc -- it's clang + C23 all the way ;-) | 17:56:20 |
| Philip Taron (UTC-8) | Modules (the clang impl) works great with C. | 17:56:39 |
| Philip Taron (UTC-8) | Maybe even better than with C++. | 17:56:50 |
| Philip Taron (UTC-8) | Actually, definitely better than with C++. | 17:57:00 |
| fzakaria | how do I go from Store to BinaryCacheStore | 21:55:59 |
| fzakaria | the config only return ref<Store> | 21:56:04 |
| fzakaria | dynamic_pointer_cast? | 21:57:08 |
| fzakaria | Sergei Zimmerman (xokdvium): writing a unit test for that issue you sent me is a bit harder because I think it's S3 specific (needs a store that uses query params)
For now I have https://github.com/NixOS/nix/pull/14266
One option: write a store just for tests that is simiilar to HttpBinaryCacheStore but sends the derivation path as a query parameter to test the encoding?
| 23:37:42 |
| Sergei Zimmerman (xokdvium) | In reply to @fzakaria:one.ems.host
Sergei Zimmerman (xokdvium): writing a unit test for that issue you sent me is a bit harder because I think it's S3 specific (needs a store that uses query params)
For now I have https://github.com/NixOS/nix/pull/14266
One option: write a store just for tests that is simiilar to HttpBinaryCacheStore but sends the derivation path as a query parameter to test the encoding?
The S3 stuff is no longer relying on custom logic (since a couple hours ago) tho | 23:38:32 |
| Sergei Zimmerman (xokdvium) | Did that maybe just fix the issue entirely? | 23:39:01 |
| fzakaria | it might have ; i included some change to src/libstore/binary-cache-store.cc that I think is "correct" but it doesn't have a unit test. | 23:39:42 |
| fzakaria | * it might have | 23:40:47 |
| fzakaria | Anyways; i pushed a unit test... took me a while to right. 🫠 | 23:41:05 |
