| 13 Feb 2024 |
 Pol | I will bisect properly from home... | 15:44:53 |
| Pol | If you want to help, I just published the shell script that I use to test | 15:47:57 |
| Pol | This is basically replicating what pkgs.php.buildComposerProject {} is doing. | 15:48:16 |
| Pol | Ok I can reproduce it locally now :) | 16:11:01 |
| Pol | I will post the issue on composer/composer ! | 16:11:08 |
| Pol | But first, I need to bisect it so I can show which commit broke it. | 16:12:03 |
| Pol | The issue is in between 2.6.6 and 2.7.0 | 16:14:06 |
| Pol | I have a hunch, a gut feeling that tells me that this might be the issue: * Fixed update --lock not keeping the dist reference/url/checksum pinned (#11787) | 16:15:44 |
| Pol | Not sure yet. | 16:15:45 |
| Pol | ❯ git bisect bad
042a8c212801aeac42b7a41b42cd1185ae28123a is the first bad commit
commit 042a8c212801aeac42b7a41b42cd1185ae28123a
Author: Jordi Boggiano <j.boggiano@seld.be>
Date: Wed Jan 10 13:33:49 2024 +0100
Ensure dist url/type/checksum remain the same when doing lock hash updates, refs #11787
src/Composer/DependencyResolver/LockTransaction.php | 3 +++
1 file changed, 3 insertions(+)
~/C/d/composer > 042a8c2(BISECTING) > php ❯
| 16:47:53 |
| Pol | Right... we have the guilty commit. | 16:48:13 |
| Pol | Now I have to make a clean proof of concept issue to submit to Composer guys | 16:48:31 |
| Pol | Job done: https://github.com/composer/composer/pull/11850 | 18:32:57 |
| Pol | Feeeew | 18:33:00 |
| Pol |
I hope there will be a good outcome of this issue | 20:41:22 |
 @patka_123:matrix.org | If I have time and energy tomorrow I'll have a look into it because it seems interesting. Good job on the work! | 20:42:55 |
| Pol | Thanks :) | 20:45:42 |
| 14 Feb 2024 |
| Pol | I hope we won't have to recalculate the hash of all PHP derivations :S | 07:43:30 |
| @patka_123:matrix.org | I hope not. But I can help with that and I guess it would just take a few hours at most | 08:03:18 |
| Pol | Yeah I might need a bit of help on this | 08:04:10 |
| Pol | I'm curious to know if the proposed fix is appropriate | 08:04:22 |
| Pol | So basically, this commit: https://github.com/composer/composer/pull/11850/commits/68e1ac0c3a2f61b8e711896bca8a40da09719e8d | 08:04:37 |
| Pol | The idea is to let composer update those values if they hasn't been changed in the meantime. | 08:11:09 |
| 15 Feb 2024 |
| Pol | I'm looking for a review here (very easy): https://github.com/NixOS/nixpkgs/pull/289146 | 21:29:09 |
| Pol | PHP bump available: https://github.com/NixOS/nixpkgs/pull/289138 | 22:26:16 |
| 16 Feb 2024 |
 Jan Tojnar | I have always done what philipp said. Having php picked up from PATH might work in theory but that breaks composer working without php on PATH (not sure how common that is) | 07:29:56 |
| Jan Tojnar | (well, we could have an executable that tries php from PATH, falling back to a hardcoded one but IME, this kiind of magic often leads to confusion) | 07:32:23 |
| 17 Feb 2024 |
 Stéphan | I wrote about dealing with private packages: https://stephank.nl/p/2024-02-17-using-nix-flake-inputs-with-php-composer.html | 20:43:41 |
| Pol | In reply to @stephank:stephank.nl
I wrote about dealing with private packages: https://stephank.nl/p/2024-02-17-using-nix-flake-inputs-with-php-composer.html Massive ! | 21:09:45 |
| @patka_123:matrix.org | Nicely written! :D | 21:42:15 |
