5 Sep 2024 |
outfoxxed | * it sounds like the resource path only, but when I compiled a module from my own application it copied the source to the module dir as well | 08:06:43 |
K900 | Here's a crazy idea | 08:07:27 |
K900 | Just make it log very loudly when it's trying to cache bare QML | 08:07:41 |
K900 | Along with the file path | 08:07:44 |
K900 | And run that for a bit | 08:07:50 |
K900 | And see when it actually gets hit, if ever | 08:07:57 |
outfoxxed | doesn't even need to be loudly | 08:07:58 |
outfoxxed | just needs a category | 08:08:01 |
K900 | Well very loudly is helpful for finding it in the journal later | 08:08:09 |
K900 | But otherwise yes | 08:08:14 |
outfoxxed | I don't run many qml applications or plasma so I can't really test that | 08:08:23 |
outfoxxed | someone with plasma would be the best choice | 08:08:29 |
K900 | I can volunteer myself but probably not today | 08:09:35 |
outfoxxed | my own application exclusively uses fs paths but they're all running through a url interceptor which makes them impossible to cache and test | 08:10:09 |
outfoxxed | What route do you think we should take the general validation storage though? I'm thinking libraryVersionHash, and I'll implement that probably tomorrow if you don't have any objections to that part | 08:18:32 |
K900 | You mean hash application build-id into the timestamp field? | 08:19:14 |
outfoxxed | no, i mean the store path into libraryVersionHash | 08:22:02 |
outfoxxed | if any of this goes upstream they can just allocate fields for it | 08:22:15 |
outfoxxed | we can't do that without increasing the maintenance burden of the patch | 08:22:44 |
outfoxxed | In reply to @k900:0upti.me I can volunteer myself but probably not today and by the way if you want to log disk cache loads, ExecutableCompilationUnit::loadCacheFilePath | 08:23:56 |
outfoxxed | Redacted or Malformed Event | 08:24:03 |
outfoxxed | you probably want to give it a custom logging category so its easy to search for in logs | 08:24:21 |
outfoxxed | that function will only be called for disk stores and lookups | 08:24:52 |
K900 | In reply to@outfoxxed:outfoxxed.me no, i mean the store path into libraryVersionHash Like, the application store path? | 08:26:41 |
K900 | That feels very cursed tbh | 08:26:49 |
outfoxxed | yes | 08:26:49 |
K900 | Though I guess if we do that we can drop the Qt cmake hack | 08:26:59 |
outfoxxed | its less cursed than anything else we discussed | 08:27:00 |
K900 | Which is also pretty cursed | 08:27:04 |
outfoxxed | i think we can drop the cmake hack regardless | 08:27:12 |