Time  Nick      Message
00:16 lhofhansl Does anybody know why EmergeManager::pushBlockEmergeData would be called with peer_request == PEER_ID_INEXISTENT and why in that case we would not enforce the disk/emerge queue limits?
00:17 lhofhansl I found that when emergequeue_limit_total is increase (by 10) the scene loads *much* faster, and the main reason is PEER_ID_INEXISTENT can load more blocks.
00:18 lhofhansl I suppose this is the server itself requesting nearby blocks?
00:27 lhofhansl Also I'd like to get #10608 in.
00:27 ShadowBot https://github.com/minetest/minetest/issues/10608 -- Revert #10495 and simplify. by lhofhansl
13:23 zughy[m]  So, I have mods crashing out of nowhere when they usually work as a charm, where the lowest common denominator are these packet quota warning messages. I know nothing about networking, but... isn't this a big problem? Server is on 5.3
13:23 zughy[m]  https://gitlab.com/zughy-friends-minetest/magic-compass/uploads/e2616653ab9b38b322ef745bdc4fc4dd/image.png
13:39 sfan5     that sounds unrelated, what does the code around that place do? what is ID?
14:07 zughy[m]  Heh, there is no easy way to explain it. A number retrieved from the name of a button clicked inside a formspec. https://gitlab.com/zughy-friends-minetest/magic-compass/-/blob/master/player_manager.lua#L15
14:08 zughy[m]  the only thing players can interact with are buttons, and every button contain a number in its name
14:14 zughy[m]  I also have another crash happened on another mod where it didn't find the player (which is kind of impossible) and an user reporting a crash where instead of ID, it didn't find "item". All these crashes can't be replicated on a local server
14:14 zughy[m]  The user I'm talking about: https://gitlab.com/zughy-friends-minetest/magic-compass/-/merge_requests/1
14:24 sfan5     >string.match on the entire fields sent by a client
14:24 sfan5     uhh
14:24 sfan5     yeah I can't think of a less reliably way of doing that
14:27 celeron55 that's amazing
14:27 celeron55 local ID = string.match(dump(fields), "%d+")
14:29 celeron55 what kind of thought process results in this
14:33 celeron55 it might be "error correcting" in way but it's also impossible to get useful log output or anything if a problem occurs
14:33 zughy[m]  lol, fine, I'll bait. The fact the formspec is configurable, so I don't know which slot is gonna contain an item, and (if it contains it) what item. Every item teleports to a certain location when clicked (linked to the ID), so I put the ID in the field name in order to retrieve it
14:34 celeron55 and the MT API could change to include a number in some field and then it'll grab that instead of your field
14:35 sfan5     chances are if you add extra debug output there you'll find that the ID it matches it not what you expected
14:35 celeron55 format the field name to include an identifier with id, like magicbutton_1337 and then detect that field and parse the id from it
14:35 sfan5     s/it/is/
14:36 celeron55 or something like that
14:36 celeron55 this might not fix your original problem but at least it's easier to debug then
14:38 celeron55 i'd say what happens currently is that the form gets closed in some way such that (fields.EMPTY or fields.quit) == false but there's no ID in any field
14:38 celeron55 and then your item is nil
14:38 celeron55 and you don't check that
14:39 celeron55 of course you could just add that check and forget about it
14:39 celeron55 that would be in the spirit of the existing code
16:26 lhofhansl Re my question yesterday about peer with id PEER_ID_INEXISTENT asking for a lot of blocks, that is related to ABMs. Loading blocks needed for ABMs competes with the loading visible blocks.
16:36 lhofhansl sfan5: are you ok with #10608?
16:36 ShadowBot https://github.com/minetest/minetest/issues/10608 -- Revert #10495 and simplify. by lhofhansl
16:42 sfan5     yes
16:43 lhofhansl Cool. Thanks! I'll merge within the hour.
16:44 lhofhansl Still thinking about block loading for ABMs. And the nothing-to-send timer to make this more smooth especially with more players.
17:08 lhofhansl Should requests for loading blocks on behalf of ABM be subject to disk and generate enqueue limits? (Currently they are not)