Time  Nick       Message
19:34 lhofhansl  Hi... Will merge #16931 soon (trivial and slight client performance change)
19:34 ShadowBot  https://github.com/luanti-org/luanti/issues/16931 -- Avoid a copy of block pointers in client timer update by lhofhansl
19:35 lhofhansl  Maybe a related change is possible on the server (in a different PR)
19:44 lhofhansl  Done
19:50 Krock      lhofhansl: why do you need the const_cast at all?
20:01 lhofhansl  The direct access to the sectors block is only allowed for const sectors. In the second loop the sectors are modified (blocks get removed) and so there it needs to be non-const.
20:03 lhofhansl  ... sector's block... :)
20:05 Krock      > This is explicitly only allowed on a const object since modifying anything while iterating is unsafe.
20:07 Krock      so if the returned container could be made constant but not its MapBlock value, that would be technically the best solution
20:07 Krock      since it addresses both concerns
20:08 lhofhansl  Yes. That would be better, indeed.
20:08 sfan5      it's casting the sector, not the block
20:09 sfan5      but I suppose the TimeOrderedMapBlock() constructor could take a const sector
20:10 lhofhansl  When I added that API while back, I forced this to be a const method for safety. I can try casting the returned collection to const so that I cannot be modified.
20:10 Krock      unfortunately "b.sect->deleteBlock(block);" cannot be made const
20:11 lhofhansl  In thise case TimeOrderedMapBlock needs a non-const sector, because it is later used to call non-const methods that modify it's block collection.
20:15 lhofhansl  Actually, no. Not the collection, but the Sector must be const. The Sector needs to be const, so that it is safe to use the returned collection. I think the API is good the way it is. (Of course there might other pointers to the same sector. but that's a different problem).
20:23 rubenwardy Anyone able to give this a second review please https://github.com/luanti-org/luanti/pull/14543
20:46 Krock      what happens if two games claim to be an alias?
20:46 Krock      ( .. of the same game?)
20:46 Krock      nvm. pkgmgr.find_by_gameid logs a warning