| 20 Dec 2025 |
 emily | well I was about to apply my time-box by asking if you can try cherry-picking the ones from that latest tag on top of our Xorg version to see which ones actually still do anything :P | 15:11:17 |
 K900 | But maybe if we can get it more normal we can not touch it again | 15:11:28 |
| emily | https://github.com/K900/nixpkgs/blob/d075ccf56d5c304ee6f7933d2eabb095bac5cd5c/pkgs/by-name/xo/xorg-server/darwin/find-cpp.patch is kinda gross, couldn't we just upstream using the Meson function to get the compiler | 15:14:57 |
| emily | I can try running it, but would be nice to get the patches sorted first | 15:15:28 |
| K900 | I couldn't get it working with cc.preprocess for whatever reason so I gave up | 15:16:19 |
| emily | https://github.com/NixOS/nixpkgs/pull/471160 is still pending btw | 16:34:34 |
 Grimmauld (any/all) | libxml? this time i don't even know what went wrong | 17:15:49 |
 Vladimír Čunát | AFAIK noone knows yet why the signing fails sometimes. | 17:25:03 |
| Vladimír Čunát | * AFAIK noone knows yet why the code signing fails sometimes. | 17:25:16 |
| Vladimír Čunát | (it succeeds but produces wrong signatures, so the binary is then unusable) | 17:25:47 |
| Vladimír Čunát | It doesn't seem really related to libxml2; seems just bad luck. | 17:26:01 |
| emily | I suspect it is an OS bug. | 18:32:28 |
| emily | because Zhaofeng Li has observed a derivation building successfully once and then breaking every time it is built again on the same machine. | 18:32:48 |
| emily | (or the reverse? I forget.) | 18:32:51 |
| emily | I think that ld64 might call into some OS API for codesigning stuff that may have some internal cache that is busted in some way, or something. | 18:33:15 |
| emily | but I haven't had the time to do a deep dive. | 18:33:24 |
| emily | I'm very tempted to blame the recent OS update on the builders for why it's suddenly breaking more often on Hydra. | 18:33:35 |
| emily | FWIW, the "signature" is just a hash | 18:33:57 |
| emily | I forget if we determined a preimage for the incorrect hash we get or not | 18:34:10 |
| emily | anyway the linked PR is just me cleaning up the rebuild hack i did on -next. | 18:34:29 |
| 21 Dec 2025 |
 Randy Eckenrode | ld64 uses libcodedirectory to parse signatures, but it does the signing itself using OpenSSL. | 01:53:19 |
| Randy Eckenrode | libcodedirectory is the bundled one plus my patch for checking whether a binary is linker-signed. | 01:53:43 |
 whispers (it/fae) | just to check, should all PRs with fixes for gcc 15 failures be targeted at staging-next, or only particularly impactful ones (and others make their way from master)? | 05:49:33 |
| K900 | Either works | 05:50:49 |
| 22 Dec 2025 |
| K900 | I fixed orcaslicer | 07:25:27 |
| K900 | https://github.com/NixOS/nixpkgs/commit/04de9bcc922ff493573be2c153fd775409243527 | 07:25:28 |
| K900 | I am going to scream | 07:25:31 |
| Grimmauld (any/all) | oh hey that probably fixes prusa slicer too | 07:59:39 |
| K900 | Yes it does | 08:01:26 |
| K900 | Very likely it fixes all the slic3r derivatives | 08:01:48 |
