| 5 Sep 2024 |
K900 ⚡️ | If we have bare QML files in the store | 07:47:59 |
outfoxxed | and no guarantee they're all qrc:/ resources either | 07:47:59 |
K900 ⚡️ | We will never clean up those | 07:48:07 |
K900 ⚡️ | Because we'll be hashing the full store path | 07:48:20 |
K900 ⚡️ | So we'll never get a wrong hit' | 07:48:25 |
K900 ⚡️ | * So we'll never get a wrong hit | 07:48:26 |
outfoxxed | hm | 07:48:42 |
outfoxxed | yeah | 07:48:44 |
outfoxxed | that was what kirigami did wasn't it | 07:48:52 |
K900 ⚡️ | I don't like this whole lookup key versus integrity key thing they've got going on | 07:48:54 |
outfoxxed | In reply to @outfoxxed:outfoxxed.me sent an image. yeah | 07:49:06 |
K900 ⚡️ | Like we're looking into the cache for one thing | 07:49:11 |
K900 ⚡️ | And then cleaning it based on something else entirely | 07:49:28 |
outfoxxed | In reply to @k900:0upti.me I don't like this whole lookup key versus integrity key thing they've got going on the purpose just seems to be so they don't spam duplicate cache files much | 07:49:32 |
K900 ⚡️ | Yeah I get that | 07:49:38 |
K900 ⚡️ | But it's making things annoying because we also can't just have tmpfilesd clean it up based on atime or whatever | 07:49:54 |
K900 ⚡️ | Because we don't know what's going on | 07:50:03 |
outfoxxed | I don't think theres a reasonable way to handle non qrc nix store qml files | 07:50:04 |
outfoxxed | we might just want to disable those entirely | 07:50:17 |
K900 ⚡️ | We could potentially base the hash off like, pname and relative path, but meh | 07:50:26 |
K900 ⚡️ | That's too cursed | 07:50:29 |
K900 ⚡️ | And too specific | 07:50:31 |
outfoxxed | wait actually | 07:50:31 |
outfoxxed | if they're in the nix store | 07:50:36 |
outfoxxed | the path will change | 07:50:40 |
outfoxxed | and thats the id key | 07:50:43 |
K900 ⚡️ | Yeah the problem is not that we'll hit stale cache | 07:50:51 |
K900 ⚡️ | The problem is that we'll never clean up stale cache | 07:50:55 |
outfoxxed | yeah i see | 07:50:59 |
K900 ⚡️ | Because it only cleans up on lookup hit | 07:51:02 |