| 5 Jul 2026 |
 K900 | https://github.com/sqlite/sqlite/blob/master/ext/qrf/qrf.c#L1137 | 21:56:32 |
| K900 | I found the source | 21:56:56 |
| K900 | We COULD patch it | 21:56:58 |
 hexa | so the pnpm fetcher gets a patched sqlite, correct? | 22:01:00 |
 emily | I mean we could really just s/X'/x'/g | 22:03:49 |
| emily | modulo escaping | 22:03:57 |
| K900 | The escaping is what I'm worried about | 22:05:23 |
| K900 | I really want someone to take this over though | 22:05:39 |
| K900 | It's 1AM | 22:06:01 |
| emily | format looks pretty uniform | 22:06:06 |
| emily | and invalidating the FODs gets you the rest of the way? | 22:06:17 |
|  Charles joined the room. | 22:08:54 |
|  mkg20001 joined the room. | 22:53:31 |
| 6 Jul 2026 |
 Vladimír Čunát | In reply to @k900:0upti.me
It's entirely unrevertable An old sqlite version for this purpose? | 05:30:40 |
| K900 | We can also just patch one character | 05:49:08 |
| K900 | But that's also not very sustainable | 05:49:20 |
|  lassulus joined the room. | 07:16:42 |
 gepbird | temporarily I'd go with pinning sqlite to an older version for pnpm only
then making a new fetcherVersion that uses can only use the new sqlite (throws with an old sqlite version) and migrate pnpm 11 packages to that
but I'm also fine with patching the FOD with sed if we can make sure it doesn't cause other issues, or reverting the sqlite change that broke it if possible | 08:30:16 |
| K900 | Reverting the change is not possible and will be a world rebuild | 08:30:33 |
| K900 | And we do not want to carry the patch going forward | 08:30:41 |
| gepbird | but long term we should replace this tooling that relies on pnpm, sqlite and other binaries, probably intercept the network requests which is what Scrumplex proposed in a PR (I haven't have time to review) | 08:31:13 |
| K900 | The correct solution is probably to figure out how to not include the SQLite data in the output at all | 08:31:14 |
| K900 | And if that's impossible, to make it so it's not represented as an internal dump format | 08:31:30 |
 Grimmauld (any/all) | seems vulnix is broken rn via zodb | 08:34:55 |
| Grimmauld (any/all) | setuptools garbage | 08:35:01 |
| K900 | If it's complaining about pkg_resources, replace setuptools with setuptools_80 | 08:35:35 |
| gepbird | for now the only solution to that seems to be https://github.com/NixOS/nixpkgs/pull/531657
but I wouldn't immediately switch all packages to this before making sure it's realiable | 08:35:49 |
| gepbird | previously it wasn't a dump format but the binary database, which has it's own can of worms | 08:36:21 |
| Grimmauld (any/all) | zodb> ERROR Missing dependencies:
zodb> setuptools<81,>=78.1.1
May be able to get away with relaxDeps | 08:39:10 |
| Grimmauld (any/all) | oh great, prusa slicer is broken | 09:37:52 |
