5 Sep 2024 |
K900 | Like OK we'll get the cache blown out once a staging cycle | 08:30:49 |
outfoxxed | alright
lets drop the patch and just write .tag in preconfigure though | 08:30:51 |
K900 | Is that really a problem | 08:30:52 |
K900 | Agreed | 08:30:55 |
outfoxxed | not as long as its able to clean up | 08:31:00 |
outfoxxed | which it can since thats validation not id | 08:31:10 |
K900 | OK so our key structure ends up being (app name, filename hash, ["build-id", mtime]) | 08:33:46 |
K900 | Where build-id is either Qt "tag" or hash(Qt "tag" + app store path) depending on whether we're outside of the store or inisde | 08:34:16 |
K900 | And when we're inside the store, mtime is 0 | 08:34:28 |
K900 | So technically we end up with either (Qt "tag", mtime) or (Qt "tag", app store path) | 08:34:51 |
K900 | And when I spell it out like that it kinda makes more sense to maybe hash the app store path into mtime instead? | 08:35:04 |
outfoxxed | the only problem with that is you have 8 bits | 08:35:19 |
K900 | Though then we don't have a way to tell if it's a hash or an actual mtime | 08:35:21 |
outfoxxed | instead of 48 bytes | 08:35:23 |
K900 | Wait is it 8 bits or 8 bytes | 08:35:28 |
outfoxxed | bytes sorry | 08:35:35 |
K900 | I feel like 8 bytes is enough honestly? | 08:35:56 |
K900 | Like we'd have to hit a different build of the same app, on the same Qt version, on the same machine | 08:36:22 |
outfoxxed | maybe, but we're already screwing with stuff right next to it and it seems like the 48 byte slot is a lot safer | 08:36:25 |
K900 | Colliding that in 8 bytes feels extremely unlikely | 08:36:36 |
outfoxxed | and lets us encode more information for raw store paths and such if we need to | 08:36:37 |
K900 | I don't really have a strong opinion one way or another | 08:37:44 |
outfoxxed | I think it just depends on how we handle bare store paths | 08:40:21 |
outfoxxed | because we might need to be able to hash more data in | 08:40:30 |
K900 | Thanks for driving this btw | 08:49:33 |
K900 | It's a good change even if getting there will probably be confusing | 08:49:48 |
outfoxxed | In reply to @k900:0upti.me Thanks for driving this btw no problem. I'm only interested because I spent a good 8 hours troubleshooting it not caching yesterday and now I want it fixed | 08:51:55 |
outfoxxed | not because I'm going to use it very much but because it annoys me | 08:52:05 |
K900 | Ain't that the spirit of open source | 08:54:29 |
outfoxxed | always | 08:58:59 |