| Time |
Nick |
Message |
| 00:05 |
|
kaeza joined #minetest-dev |
| 00:17 |
|
est31 joined #minetest-dev |
| 00:18 |
est31 |
@ shader discussion, copying the shaders before RBA's recent changes and switching back and forth shouldn't be too hard |
| 00:19 |
VanessaE |
that sounds fair |
| 00:19 |
VanessaE |
will that mess with that "tangent space" stuff he's talked about recently? |
| 00:19 |
VanessaE |
(I assume not) |
| 00:20 |
est31 |
if you replace the whole files, no |
| 00:21 |
|
sloantothebone joined #minetest-dev |
| 00:30 |
hmmmm |
est31: sure, am aware, but they aren't really too relevant |
| 00:31 |
hmmmm |
est31: will look at your PR |
| 00:31 |
hmmmm |
about shaders |
| 00:31 |
hmmmm |
games typically have one shader for each effect |
| 00:32 |
hmmmm |
the way RBA crams them all into a single shader is kind of ... |
| 00:32 |
hmmmm |
not good |
| 00:32 |
VanessaE |
bad. |
| 00:32 |
hmmmm |
so any slowness added for one effect has effects on everything else |
| 00:32 |
hmmmm |
when I start doing shader work I do intend on fixing this and making a way more modular setup for shaders |
| 00:33 |
hmmmm |
also another quick thing we could do before release is run the shader optimizer on this |
| 00:34 |
VanessaE |
there's a tool for that? |
| 00:35 |
hmmmm |
have a look at this http://aras-p.info/blog/2010/09/29/glsl-optimizer/ |
| 00:35 |
VanessaE |
huh. TIL. |
| 00:35 |
hmmmm |
oh god, reddit speak |
| 00:36 |
VanessaE |
heh |
| 00:36 |
VanessaE |
could be worse. I could be jabbering in txtspk :P |
| 00:46 |
est31 |
hmmmm, can you make the next larger thing you work on be client scripting? |
| 00:46 |
hmmmm |
that was the plan |
| 00:46 |
est31 |
good |
| 01:00 |
|
paramat joined #minetest-dev |
| 01:02 |
paramat |
will push #3061 soon |
| 01:02 |
ShadowBot |
https://github.com/minetest/minetest/issues/3061 -- Defaultsettings: Increase client_mapblock_limit to 5000 by paramat |
| 01:02 |
|
Darcidride joined #minetest-dev |
| 01:02 |
VanessaE |
paramat: see what I said about that? (it was a joke :P ) |
| 01:02 |
hmmmm |
i honestly think 5000 is too small as well |
| 01:02 |
hmmmm |
i think the default should be unlimited |
| 01:03 |
VanessaE |
well, how much mem would 10k take? |
| 01:03 |
hmmmm |
not really much |
| 01:03 |
hmmmm |
also you guys are underestimating how small a mapblock really is |
| 01:03 |
hmmmm |
like paramat said before, an 80x80x80 area is already 125 of those |
| 01:03 |
hmmmm |
think about how many mapblocks are air and have no mesh |
| 01:03 |
VanessaE |
indeed. |
| 01:03 |
hmmmm |
those count against it too |
| 01:04 |
hmmmm |
also just wondering |
| 01:04 |
hmmmm |
this limit is client-side only |
| 01:04 |
VanessaE |
right |
| 01:04 |
hmmmm |
how does the server know that the client dropped the mapblock in question? |
| 01:04 |
est31 |
yes it does |
| 01:04 |
est31 |
its sent to the server |
| 01:04 |
hmmmm |
that is |
| 01:04 |
hmmmm |
what is* |
| 01:05 |
VanessaE |
my joke btw was that, since cubert(5000) is a bit over 17, I offered up the cubes of 16, 17, and 20 in case someone complains that 5000 is not a perfect cube :) |
| 01:05 |
est31 |
hmmmm, you agree to the client leaking then? |
| 01:05 |
VanessaE |
(4096, 4913, or 8000) |
| 01:05 |
hmmmm |
est31: ??? |
| 01:05 |
est31 |
how would it be if the server never unloaded blocks |
| 01:05 |
paramat |
ah |
| 01:05 |
hmmmm |
how is it "sent to the server" though |
| 01:05 |
hmmmm |
what packet does this use |
| 01:06 |
VanessaE |
but that said, something like 15625 (cube of 25) might be usable, if slightly excessive. |
| 01:06 |
est31 |
hmmmm, https://github.com/minetest/minetest/blob/a8e238ed06ee8285ed4459e9deda3117419837f6/src/client.cpp#L428-L458 |
| 01:07 |
est31 |
TOSERVER_DELETEDBLOCKS |
| 01:07 |
hmmmm |
okay, good |
| 01:07 |
hmmmm |
just wondering |
| 01:10 |
paramat |
i'll leave that PR to ferment a little longer then |
| 01:10 |
hmmmm |
what do you think |
| 01:10 |
hmmmm |
imho i think the default should be unlimited |
| 01:11 |
VanessaE |
isn't it already? or rather, wasn't it, before the 1000 went in? |
| 01:12 |
VanessaE |
the problem with making it unlimited is that mem usage just grows and grows, unless you timestamp the loaded blocks and just throw out the oldest stuff after some amount of time |
| 01:12 |
paramat |
unlimited means the original problem of excessive memory use |
| 01:12 |
est31 |
basically, a leak |
| 01:12 |
VanessaE |
yep |
| 01:12 |
VanessaE |
well not *really* a leak |
| 01:12 |
hmmmm |
it is definitely not a leak |
| 01:12 |
VanessaE |
but the effect is the same. |
| 01:13 |
hmmmm |
if you continuously use every loaded block the effect is the same |
| 01:13 |
est31 |
its keeping around data not needed anymore |
| 01:13 |
hmmmm |
it doesn't, after a set amount of time it frees the unneeded data |
| 01:13 |
est31 |
it isn't technically a leak |
| 01:13 |
est31 |
but it is a leak the same way a lua error which stops the server process is a crash |
| 01:14 |
hmmmm |
so then reduce the default block unload timeout |
| 01:14 |
|
proller joined #minetest-dev |
| 01:15 |
hmmmm |
i recognize the new unload method as a decent way to keep memory usage below a certain threshhold. but I would not say that things were broken before this |
| 01:15 |
VanessaE |
I thought the existing timeout just didn't work? |
| 01:15 |
VanessaE |
(hence my previous comment) |
| 01:15 |
hmmmm |
if it just didn't work, then we'd fix that bug |
| 01:15 |
est31 |
it did work |
| 01:15 |
hmmmm |
not ignore it and add a new method |
| 01:15 |
est31 |
yes |
| 01:15 |
est31 |
agreed |
| 01:16 |
est31 |
but the new method is better |
| 01:16 |
paramat |
the timeout is not a good way to limit memory use because it doesn't take into account how fast a player is travelling, for example constant flying |
| 01:16 |
est31 |
^ |
| 01:16 |
VanessaE |
paramat: in a way it kinda does. |
| 01:16 |
hmmmm |
but the thing is |
| 01:16 |
est31 |
its a bad way just like taking into account where a player is right now |
| 01:16 |
VanessaE |
the faster you travel, the less time you spend in a given area, ergo the less stuff that should (theoretically) load for those areas. |
| 01:16 |
hmmmm |
this doesn't actually keep memory usage below a threshhold |
| 01:17 |
hmmmm |
just adds a theoretical maximum |
| 01:17 |
est31 |
its a rough estimate |
| 01:17 |
est31 |
yes |
| 01:17 |
hmmmm |
there could be a really nasty mapblock made up of a checkerboard pattern of solid nodes and non-solid nodes |
| 01:17 |
est31 |
yea |
| 01:17 |
hmmmm |
which would theoretically take up the most mesh space possible |
| 01:17 |
est31 |
but still, better than a timeout |
| 01:20 |
est31 |
(the timeout makes total sense for servers, btw) |
| 01:21 |
est31 |
for servers you limit player count |
| 01:21 |
est31 |
not how long a player can play |
| 01:24 |
est31 |
hmmm, feel free to add RAM useage measurement to mapblock storage |
| 01:24 |
est31 |
would be a great addition |
| 01:25 |
est31 |
what do people think should the shader regression issue be addressed before 0.4.13 |
| 01:25 |
est31 |
or should we wait until we find a new shader maintainer |
| 01:25 |
est31 |
I'm neutral about this |
| 01:29 |
est31 |
and about which default value to use, we should make it based on measurements |
| 01:30 |
est31 |
If 100k blocks take less than 500 MB in total for the whole clients, I'm ok with a limit of 100k |
| 01:30 |
est31 |
client* |
| 01:32 |
paramat |
i'm flying fast over mgv7 right now, with the 5000 value, total memory use due to mt has stabilised at 1.4GB |
| 01:32 |
hmmmm |
i think it would be pretty hard to calculate the memory usage of a mapblock since half of it is stored in irrlicht |
| 01:33 |
hmmmm |
we don't have to have a shader "maintainer" - if you are able to fix the shaders, then do so |
| 01:33 |
est31 |
this small patch makes you benchmark the number of loaded blocks : https://gist.github.com/anonymous/ef351c1b9d8222dead32 |
| 01:33 |
hmmmm |
why do people think they need to be pidgeonholed into a certain piece of code? |
| 01:36 |
est31 |
s/shader maintainer/somebody who fixes this/ |
| 01:36 |
est31 |
bye! |
| 01:38 |
VanessaE |
paramat: that's... rather a lot. |
| 01:38 |
paramat |
now i exit to menu, memory use drops to 1.3GB, it should frop further surely |
| 01:38 |
paramat |
(drop) |
| 01:38 |
VanessaE |
what's the measure on a fresh start? |
| 01:39 |
VanessaE |
1.3 GB is a fuckton |
| 01:39 |
paramat |
0.7GB was initial |
| 01:39 |
hmmmm |
yeah I agree |
| 01:39 |
hmmmm |
there's something else going on taking up too much memory |
| 01:39 |
VanessaE |
so wait, it's hanging onto 600 MB after shutdown? |
| 01:40 |
hmmmm |
we know from massif that a lot of it is noise buffers and mapblocks |
| 01:40 |
hmmmm |
but there's no way noise takes up 700MB |
| 01:40 |
paramat |
here i'm using my 'system monitor' watching total memory use |
| 01:41 |
paramat |
700MB before starting mt |
| 01:42 |
paramat |
flying fast over mgv7 rises to 2.1GB and stabilises (with setting 5000) |
| 01:42 |
paramat |
exit to menu falls to 2.0GB ? |
| 01:42 |
hmmmm |
were you using v6 for the first test? |
| 01:42 |
VanessaE |
for me, it's talking 75 MB at start, about 250MB after the world loads, 343 after it stabilizes, before I start exploring. |
| 01:42 |
paramat |
no mgv7 |
| 01:42 |
VanessaE |
(vanilla mt_game, mgv6) |
| 01:43 |
paramat |
so 1.4GB for 5000 mapblocks, seems excessive? |
| 01:43 |
VanessaE |
rises to around 720 MB if I just pan around without walking from the spot I spawned at. 403 MB if I exit to menu |
| 01:44 |
VanessaE |
so for me, it's hanging onto over 300 MB. |
| 01:44 |
paramat |
obviously something isn't being cleared when exiting to menu |
| 01:45 |
paramat |
but from previous tests the amount cleared when exiting to menu varies from almost all to almost nothing |
| 01:45 |
paramat |
oh and i have a timeout of 3 mins |
| 01:46 |
VanessaE |
correction, it's around 410 MB, not 720 (that was X, got it mixed up). |
| 01:50 |
|
proller joined #minetest-dev |
| 02:02 |
paramat |
now with limit 1000 mapblocks, mgv7 singleplayer, total mt memory use stabilises at 0.7GB. exit to menu, falls to 0.2GB, more cleared this time |
| 02:05 |
paramat |
i should increase my 'client unload timeout' to neutralise it's effect.. |
| 02:23 |
paramat |
.. not much difference. mgv7 singleplayer, mapblock limit 5000. stabilises at 1.5GB, exit to menu falls to 1.2GB |
| 02:24 |
paramat |
re-enter world, falls to 0.8GB |
| 02:28 |
paramat |
start flying fast again, rises unusually fast to 2.1GB and stabilises |
| 02:30 |
paramat |
exit to menu again, falls to 1.4GB |
| 02:41 |
paramat |
so for this release i think 5000 is about right, uses up to 2GB if re-entering worlds |
| 02:48 |
|
paramat left #minetest-dev |
| 02:55 |
VanessaE |
bbl |
| 04:43 |
|
Hunterz joined #minetest-dev |
| 06:01 |
|
Hunterz joined #minetest-dev |
| 06:11 |
|
nrzkt joined #minetest-dev |
| 06:21 |
hmmmm |
hmm |
| 06:22 |
hmmmm |
if a block gets force unloaded, should it be removed from the active block list, or should i allow the environment step to remove it? |
| 06:25 |
nrzkt |
do the two actions together ? |
| 06:27 |
|
Krock joined #minetest-dev |
| 06:40 |
|
Hunterz joined #minetest-dev |
| 06:42 |
nrzkt |
will push #3054 |
| 06:42 |
ShadowBot |
https://github.com/minetest/minetest/issues/3054 -- Remove unused function getEvent from connection.{cpp,h} by nerzhul |
| 07:20 |
|
leat joined #minetest-dev |
| 07:25 |
VanessaE |
hmmmm: still here? |
| 07:29 |
hmmmm |
yeah? |
| 07:31 |
VanessaE |
just reminding that #1425 needs fixed before release also |
| 07:31 |
ShadowBot |
https://github.com/minetest/minetest/issues/1425 -- Deleting extraneous entities from a mapblock deletes PLAYERS also |
| 07:31 |
VanessaE |
and whatever else is responsible for causing players to sometimes "lose contact" after connecting to VE-Creative |
| 07:31 |
|
nore joined #minetest-dev |
| 07:32 |
hmmmm |
wtf how old is this bug |
| 07:32 |
VanessaE |
over a year. |
| 07:33 |
VanessaE |
and this whole shit about not being able to safely run rollback because there's no way to selectively keep events out of it |
| 07:34 |
hmmmm |
so aside from clearobjects, how else are extraneous entities deleted? |
| 07:34 |
VanessaE |
(think for example a machine that's sitting there placing and digging stuff, like an automated tree farm. bloats the rollback log into the 10's of gigs) |
| 07:35 |
VanessaE |
they're allegedly deleted by the "too many objects" code. |
| 07:36 |
hmmmm |
well i don't think it's necessarily bad to delete playersaos |
| 07:36 |
VanessaE |
why? |
| 07:36 |
VanessaE |
doing so disconnects the player for no good reason |
| 07:36 |
hmmmm |
exactly, that shouldn't be the case |
| 07:36 |
hmmmm |
rather no |
| 07:36 |
hmmmm |
I mean it's not bad to delete statically stored playersaos |
| 07:37 |
hmmmm |
i mean what kind of data do they have? their data should all be in worlds/<world>/players/* |
| 07:38 |
hmmmm |
lol statically stored playersao is an oxymoron |
| 07:38 |
hmmmm |
player static objects should be safely deletable |
| 07:38 |
hmmmm |
playersaos should never be deleted on the other hand |
| 07:39 |
hmmmm |
deactivateFarObjects() is a total clusterfuck |
| 07:40 |
hmmmm |
it's too big, the logic is too complex and fragile |
| 07:42 |
hmmmm |
hmm |
| 07:42 |
hmmmm |
are playersaos even allowed to have a static object counterpart?? |
| 07:42 |
hmmmm |
indeed no |
| 07:43 |
hmmmm |
deactivateFarObjects() is player-safe. same with clearAllObjects() |
| 07:47 |
hmmmm |
kahrl fixed the latter with this https://github.com/minetest/minetest/commit/88cdd3a363668cfb2fd07b5381cce29738aed0bf |
| 07:54 |
hmmmm |
VanessaE: what attempt to fix the bug are you talking about exactly? |
| 07:55 |
VanessaE |
I don't remember what that was referring to |
| 07:56 |
VanessaE |
probabvly the situation where //clearobjects (two slashes, worldedit) could also delete player objects. |
| 07:56 |
VanessaE |
probably* |
| 07:57 |
hmmmm |
lol, no... the fix for that was made in early 2012 |
| 07:57 |
hmmmm |
er wait a minute |
| 07:57 |
VanessaE |
eh... no |
| 07:57 |
hmmmm |
why does worldedit have its own separate clearobjects command |
| 07:57 |
VanessaE |
that's for clearing objects just in the worldedit-selected region. |
| 07:57 |
VanessaE |
most of this started with ShadowNinja's SAO rewrite a while back. |
| 07:59 |
hmmmm |
https://github.com/Uberi/Minetest-WorldEdit/blob/163dffccb38a1f4e37866d7232cecfb5238f7af7/worldedit/manipulations.lua#L583 |
| 07:59 |
hmmmm |
???? |
| 08:00 |
|
Yepoleb_ joined #minetest-dev |
| 08:00 |
|
Akuma joined #minetest-dev |
| 08:00 |
VanessaE |
not sure when that check was added. probably around the same time as the "attempt" in the engine. |
| 08:02 |
hmmmm |
do you still see this problem regularly? |
| 08:02 |
VanessaE |
you mean signing on and getting no map data/player deleted on signon? |
| 08:02 |
hmmmm |
that?? or 1425 |
| 08:02 |
hmmmm |
cause i was talking about 1425 |
| 08:03 |
VanessaE |
the same. happened just a few mins before I mentioned it initially. |
| 08:03 |
hmmmm |
welll |
| 08:03 |
hmmmm |
how are you sure they're the same |
| 08:03 |
VanessaE |
I'm not sure, but the pattern is the same as it's been since I reported that bug |
| 08:03 |
VanessaE |
VE-Creative is difficult to sign onto to just play |
| 08:04 |
VanessaE |
because as soon as you're in, somewhere the engine thinks there are too many entities, deletes them, and deletes you. |
| 08:05 |
hmmmm |
so do you see Cancelling: No player for peer_id=# in the log at the same time as the player attempts to log in but cannot get map? |
| 08:05 |
hmmmm |
if you do, then it's the same bug |
| 08:05 |
hmmmm |
if not, it's a different thing |
| 08:05 |
hmmmm |
the engine thinking there are too many entities simply does not apply to players |
| 08:05 |
hmmmm |
player objects are not allowed static counterparts so they're skipped over in deactivateFarObjects |
| 08:06 |
hmmmm |
sorry to blow a huge hole in your theory :p |
| 08:07 |
VanessaE |
I *do* see the "Cancelling" message |
| 08:07 |
VanessaE |
right after I joined, in fact. |
| 08:07 |
hmmmm |
aren't you supposed to get disconnected? |
| 08:07 |
VanessaE |
sorry to blow a hole in your explosives ;) |
| 08:08 |
VanessaE |
here's a log excerpt from that last event: http://pastebin.com/SRzBmb9a |
| 08:08 |
VanessaE |
note line 18... the per-block limit on this server is 500. |
| 08:09 |
hmmmm |
did you really emerge at block (-43, 1, -28)? |
| 08:11 |
VanessaE |
probably. I was off somewhere I didn't recognize. |
| 08:11 |
hmmmm |
well |
| 08:11 |
VanessaE |
wasn't the usual spawn area but some road and house, probably something cheapie had showed me recently. |
| 08:11 |
hmmmm |
are you sure there aren't any mods that spawn objects right next to you as you join the game |
| 08:11 |
hmmmm |
because that's what I think is going on |
| 08:11 |
VanessaE |
I don't see how |
| 08:11 |
hmmmm |
i'm telling you, players never get deactivated in the routine that error message originates from |
| 08:12 |
hmmmm |
simply not possible |
| 08:12 |
VanessaE |
I've seen it happen in areas with nothing whatsoever that could cause an object to be created. |
| 08:13 |
hmmmm |
so as it stands right now I understand this bug as: |
| 08:13 |
hmmmm |
"Player seems to successfully emerge but object gets deleted soon after" |
| 08:13 |
VanessaE |
yep |
| 08:13 |
VanessaE |
and if the player *does* successfully emerge and can hang about for more than a few seconds, then they generally don't get deleted after that |
| 08:14 |
VanessaE |
it's like there's a several-second window right after emerge where they're vulnerable. |
| 08:16 |
hmmmm |
i'd love to check it out |
| 08:16 |
hmmmm |
but it's 4:15 AM |
| 08:16 |
VanessaE |
I know. |
| 08:16 |
VanessaE |
just point this out becasue this REALLY needs fixed before 0.4.12 |
| 08:16 |
VanessaE |
er 0.4.13 |
| 08:16 |
VanessaE |
as does tha rollback issue |
| 08:17 |
VanessaE |
there NEEDS to be some way to prevent rollback from logging an action so that automated machines don't bloat the log |
| 08:20 |
hmmmm |
then make a jira issue about it |
| 08:20 |
hmmmm |
er, github issue |
| 08:20 |
VanessaE |
I think such already exists., |
| 08:20 |
VanessaE |
https://github.com/minetest/minetest/issues/2335 |
| 08:24 |
VanessaE |
as you can see, no one's done anything about it. |
| 08:24 |
VanessaE |
right now I'm cleaning up, as c55 once put it, "a fuckton of griefing" by hand because of that. |
| 08:25 |
hmmmm |
i think rollback sucks |
| 08:25 |
hmmmm |
snapshots should be made instead |
| 08:26 |
hmmmm |
diff-based snapshots |
| 08:28 |
VanessaE |
wouldn't work so well in realtime |
| 08:28 |
VanessaE |
rollback's main purpose is to undo griefing |
| 08:28 |
VanessaE |
which needs at least per-minute granularity |
| 08:31 |
|
kilbith joined #minetest-dev |
| 08:50 |
VanessaE |
paramat: aliases are needed for pine wood slabs/stairs. |
| 09:18 |
|
FR^2 joined #minetest-dev |
| 09:40 |
VanessaE |
hmmmm: here's the latest OOM crash, now with modname, callback, and mem usage: http://pastebin.com/KHETC1Wa (already pinged Sokomine, markers is hers) |
| 09:53 |
|
kilbith left #minetest-dev |
| 10:05 |
|
WSDguy2014 joined #minetest-dev |
| 10:11 |
|
WSDguy2014 left #minetest-dev |
| 10:17 |
|
srifqi joined #minetest-dev |
| 10:37 |
|
proller joined #minetest-dev |
| 11:13 |
|
Zeitgeist_ joined #minetest-dev |
| 11:16 |
kahrl |
VanessaE: is your version of gdb by any chance incompatible with your compiler version? |
| 11:16 |
kahrl |
because I see lots of dwarf errors there |
| 11:29 |
|
leat joined #minetest-dev |
| 11:54 |
sfan5 |
https://github.com/minetest/minetest_game/pull/627#issuecomment-131355267 |
| 11:54 |
sfan5 |
>Poorly contrasted greenish planks VS. Faithful well-aged planks... but, well, as long as paramat is actively lobbying his stuff, it'll be preferred for paramat_game and never declined, as usual. No wonder why people like BlockMen are left (including me right now). This game is lost. |
| 11:54 |
sfan5 |
lolwat |
| 12:05 |
|
kilbith joined #minetest-dev |
| 12:05 |
kilbith |
ah sfan5 canalizing the attention of the pack to take offense against me, basely in my back |
| 12:06 |
kilbith |
you're a coward guy that only wake up when the collective is/may against me |
| 12:06 |
sfan5 |
correct, i never do actual development work on minetest |
| 12:06 |
sfan5 |
it's obvious isn't it? |
| 12:06 |
kilbith |
where did i said that ? |
| 12:07 |
sfan5 |
it's kind of implied |
| 12:07 |
kilbith |
it's interpreted and farfetched |
| 12:07 |
kilbith |
so lay off |
| 12:08 |
kilbith |
if you have some kind of problem against me, tell that when i can ensure my defense |
| 12:08 |
kilbith |
otherwise you're just a coward |
| 12:08 |
sfan5 |
<kilbith> it's interpreted and farfetched |
| 12:08 |
sfan5 |
are you telling me what you commented on github is not interpreted and farfetched? |
| 12:09 |
kilbith |
about what in particular ? |
| 12:09 |
sfan5 |
https://github.com/minetest/minetest_game/pull/627#issuecomment-131355267 this one |
| 12:10 |
kilbith |
i know but what bit of that comment offend you |
| 12:10 |
sfan5 |
where did i say that |
| 12:10 |
kilbith |
"i never do actual development work on minetest" <- where i implied that ? |
| 12:11 |
kilbith |
it's just offtopic and out of the target |
| 12:11 |
sfan5 |
"it's just offtopic" |
| 12:11 |
sfan5 |
you're funny |
| 12:12 |
kilbith |
for the umpteenth time you just wanted to present me as a villain |
| 12:12 |
sfan5 |
i don't even know what you problem is |
| 12:12 |
sfan5 |
all i did was comment on your github comment with "lolwat" |
| 12:12 |
kilbith |
anyways, i don't care anymore |
| 12:13 |
sfan5 |
then you go accusing me of only trying to take offense against you |
| 12:13 |
sfan5 |
ah whatever |
| 12:14 |
|
VargaD joined #minetest-dev |
| 12:23 |
|
SopaXorzTaker joined #minetest-dev |
| 13:09 |
Krock |
as usually, there's no pull request that someone doesn't like |
| 13:12 |
|
H-H-H joined #minetest-dev |
| 14:43 |
|
Gael-de-Sailly joined #minetest-dev |
| 14:44 |
|
hmmmm joined #minetest-dev |
| 14:46 |
VanessaE |
kahrl: well I assume gdb and gcc/g++ are compatible with one another -- both are the packages that come with xubuntu. |
| 15:08 |
|
RealBadAngel joined #minetest-dev |
| 15:09 |
RealBadAngel |
hi |
| 15:09 |
RealBadAngel |
hmmm youre quite wrong about that shaders optimizer |
| 15:09 |
RealBadAngel |
this is not meant for open gl just for ES |
| 15:10 |
RealBadAngel |
so good luck using that. you will just make shaders human unreadable |
| 15:10 |
|
Zeitgeist_ joined #minetest-dev |
| 15:11 |
RealBadAngel |
other thing. theres no one shader for everything |
| 15:11 |
RealBadAngel |
shaders are instanced for each drawtype |
| 15:12 |
RealBadAngel |
as usual you have ideas to make the code better but not a slightest idea how it works |
| 15:13 |
RealBadAngel |
dont make an idiot of yourself and keep working on areas you know |
| 15:45 |
|
leat joined #minetest-dev |
| 15:45 |
hmmmm |
so you came around just to taunt me? |
| 15:45 |
hmmmm |
don't become a proller |
| 15:48 |
RealBadAngel |
no |
| 15:48 |
|
Player_2 joined #minetest-dev |
| 15:48 |
RealBadAngel |
i am answering to what you have said while i was absent |
| 15:49 |
RealBadAngel |
you propably have never read the link about optimizer you have put there |
| 15:49 |
|
tuy joined #minetest-dev |
| 15:49 |
RealBadAngel |
otherwise you want be saying such bullshits |
| 15:49 |
hmmmm |
look, fuck off back to your fork or whatever already, i'm tired of your attitude |
| 15:50 |
hmmmm |
maybe it was for the best that you left minetest |
| 15:50 |
RealBadAngel |
so stop insulting me behind my back |
| 15:50 |
RealBadAngel |
there are logs, i can still read them :P |
| 15:51 |
hmmmm |
i never said anything insulting about you |
| 15:51 |
hmmmm |
never mind that. it's off topic for this channel. |
| 15:51 |
RealBadAngel |
i also never mind that. that part is over for me |
| 15:52 |
RealBadAngel |
so just please read what you have posted and check out how shaders work |
| 15:53 |
|
ElectronLibre joined #minetest-dev |
| 16:00 |
VanessaE |
regardless of who knows what about which feature, we need simple shaders equivalent to what was present just after the "Waving" stuff went in, before 0.4.13 goes out. |
| 16:09 |
hmmmm |
if we do that, is there any point to having shaders enabled? |
| 16:09 |
hmmmm |
is the acceleration of finalColorBlend actually worth it? |
| 16:09 |
VanessaE |
idk |
| 16:09 |
VanessaE |
for me, shaders cost fps. |
| 16:10 |
hmmmm |
then the answer is no |
| 16:10 |
VanessaE |
for others, evidently they're faster |
| 16:10 |
hmmmm |
fwiw I run with shaders disabled anymore, I still take a nasty FPS hit since that one commit |
| 16:10 |
RealBadAngel |
then stay in stone age |
| 16:10 |
VanessaE |
RealBadAngel: R9 280X is NOT "stone age". |
| 16:10 |
VanessaE |
if shaders are slower than baseline code on my GPU, there's something wrong with the shaders code. |
| 16:11 |
RealBadAngel |
my gtx also is not |
| 16:11 |
RealBadAngel |
and yet i do get twice or four times more fps than you or hmm |
| 16:12 |
RealBadAngel |
with all the possible effects around |
| 16:13 |
RealBadAngel |
more i let to be done GPU side, more fps i get |
| 16:14 |
RealBadAngel |
also i cannot see any difference with shaders enabled and disabled |
| 16:14 |
RealBadAngel |
i just get more fancy gfx with shaders |
| 16:14 |
RealBadAngel |
disabling shaders at all doesnt cause any speed up as you are claiming |
| 16:15 |
RealBadAngel |
just world lighting is fucked up by definition |
| 16:16 |
RealBadAngel |
anyway lite shaders will use glsl 2.1 enough for older boxes |
| 16:16 |
VanessaE |
kaeza uses 1.2... |
| 16:17 |
|
paramat joined #minetest-dev |
| 16:17 |
RealBadAngel |
we are mixing glsl and opengl versions |
| 16:18 |
VanessaE |
Irrlicht log: GLSL version: 1.2 |
| 16:18 |
VanessaE |
you were saying? |
| 16:18 |
VanessaE |
https://github.com/minetest/minetest/issues/2866 |
| 16:18 |
RealBadAngel |
https://en.wikipedia.org/wiki/OpenGL_Shading_Language |
| 16:19 |
RealBadAngel |
OpenGL 2.1 comes with shading language GLSL version 1.2 |
| 16:20 |
VanessaE |
"<RealBadAngel> anyway lite shaders will use glsl 2.1 enough for older boxes" |
| 16:20 |
RealBadAngel |
#version 120 will be absolute minimum to talk about shaders |
| 16:20 |
VanessaE |
"we" are not mixing anything up. you are. |
| 16:20 |
RealBadAngel |
current version is 450 |
| 16:20 |
RealBadAngel |
and this one i will support and use with regular shaders |
| 16:21 |
RealBadAngel |
yeah, i mixed that. |
| 16:22 |
RealBadAngel |
versions #120 is the correct one for GLSL |
| 16:22 |
ShadowBot |
https://github.com/minetest/minetest/issues/120 -- lua_api.txt is missing important stuff |
| 16:23 |
RealBadAngel |
120 is several years old |
| 16:23 |
RealBadAngel |
enough even for watches to work with |
| 16:24 |
RealBadAngel |
but dont expect even bumpmapping working there properly |
| 16:24 |
RealBadAngel |
just final color blend, some waving and period |
| 16:25 |
VanessaE |
I think that's all that's wanted. |
| 16:25 |
VanessaE |
no one expects fancy shaders like that to work on old GPUs |
| 16:26 |
VanessaE |
but a simple shader like waving or lighting should work everywhere |
| 16:26 |
paramat |
nore sfan5 your opinions on game#631 ? |
| 16:26 |
ShadowBot |
https://github.com/minetest/minetest_game/issues/631 -- Default/functions: Half rate of leafdecay ABM by paramat |
| 16:26 |
hmmmm |
speaking of waving we need to draw backfaces on waving nodes |
| 16:26 |
nore |
hm... I would agree with est31 on this one |
| 16:27 |
paramat |
yeah it's controversial =) |
| 16:27 |
RealBadAngel |
hmmmm, not only them. |
| 16:28 |
RealBadAngel |
but this i want to move to geometry shaders at all |
| 16:28 |
RealBadAngel |
thats a proper place to do such things |
| 16:28 |
RealBadAngel |
including extruding things |
| 16:32 |
RealBadAngel |
meanwhile some1 could took care of guys flying around on hacked clients and destroyin other players work |
| 16:32 |
RealBadAngel |
that should be more important for you than a few fps |
| 16:34 |
RealBadAngel |
while my work can be turned off you cant turn on a guy flying around and watchin while he is destroyin what you have build for weeks |
| 16:34 |
RealBadAngel |
*turn off |
| 16:36 |
paramat |
hmmmm okay for #3061 ? from my testing in logs seems okay for now |
| 16:36 |
ShadowBot |
https://github.com/minetest/minetest/issues/3061 -- Defaultsettings: Increase client_mapblock_limit to 5000 by paramat |
| 16:36 |
hmmmm |
I guess it's okay |
| 16:36 |
paramat |
okay will push later |
| 16:38 |
|
Dragonop joined #minetest-dev |
| 16:38 |
Krock |
paramat, why 5k? |
| 16:38 |
paramat |
see discussion linked to |
| 16:39 |
|
Dragonop_ joined #minetest-dev |
| 16:39 |
Krock |
oh. Haven't seen that link |
| 16:40 |
Krock |
gonna change it to 200 when I got a new build :P |
| 16:40 |
paramat |
that's way too low, not even 2 mapchunks |
| 16:41 |
Krock |
mapchunks != mapblocks?? |
| 16:41 |
paramat |
1 mapchunk = 125 mapblocks |
| 16:42 |
Krock |
oh |
| 16:43 |
paramat |
or rather 1 chunk contains 125 blocks |
| 16:43 |
paramat |
5^3 |
| 16:44 |
Krock |
but the chunk size is changeable |
| 16:44 |
paramat |
yeah, 125 with the default |
| 16:45 |
paramat |
mapblocks are fixed size, so the setting is in mapblocks |
| 16:49 |
|
Calinou joined #minetest-dev |
| 16:50 |
|
Dragonop joined #minetest-dev |
| 16:56 |
Taoki |
hmmmm, RealBadAngel: I can confirm that disabling shaders increases FPS very noticeable, actually. This is of course understandable, shaders cost the GPU. |
| 16:57 |
Taoki |
The slowdown isn't much more dramatic than say, when enabling shaders in a Quake engine fork. In Minetest it might be a bit slower though, since there might be a bit more unoptimized surfaces (being a voxel engine and all) |
| 16:59 |
Taoki |
In any case, RealBadAngel has done an amazing work with shaders since he's been working on them. Which is one reason why I'm among those wanting him to stay with Minetest (that doesn't impede him working on his own fork as well, pluu requsts should be easy to make to both projects). |
| 16:59 |
Calinou |
most games today don't offer fixed-function fallbacks because it's a huge code saver; Sauerbraten and Red Eclipse recently removed fixed-function rendering support, bumping the minimal OpenGL version supported to 2.1 (2006-2007 era) |
| 17:00 |
Taoki |
OpenGL 2.1 is a very reasonable requirement in my book. |
| 17:00 |
Calinou |
yes, we shouldn't be afraid to require that |
| 17:00 |
Calinou |
this is not OpenArena (I hope) |
| 17:00 |
Taoki |
Not like what High Fidelity did... bumping the minimum required version to OpenGL 4.1, which is why I'm not unable to use it any more. |
| 17:01 |
Calinou |
the free drivers will soon support OpenGL 4.1 |
| 17:01 |
Calinou |
but it'll be a year or so until it lands in distributions |
| 17:01 |
Taoki |
Calinou: There has been a LOT of conflict in the Minetest dev team over new vs. old. Some (like RBA) want to sacrifice very old compatibility for newer features. Others want Minetest to stay compatible with computers 20 years old. |
| 17:01 |
Calinou |
I don't think it's compatible with 1995 computers :) |
| 17:01 |
Taoki |
And of course, not many make an effort to stay civil about it either it appears :/ |
| 17:02 |
Taoki |
so hopefully this is something that can be dissected better someday. |
| 17:02 |
Calinou |
we need more AAA game competitors, and for that we need better graphics, it has been proven it's the only way to attract people |
| 17:02 |
Taoki |
That's surprising :P |
| 17:02 |
Taoki |
Calinou: +1 |
| 17:03 |
Taoki |
Better graphics TOO, not ONLY better graphics. Otherwise we end up like Unreal Tournament 3: Great graphics, the most depressive and useless gameplay. |
| 17:03 |
Taoki |
Thankfully that's not the case with Minetest, the gameplay and functionality is great already. |
| 17:03 |
Calinou |
but the stability isn't :s |
| 17:03 |
Taoki |
True |
| 17:03 |
Calinou |
(well, many AAA games are poorly optimized and unstable too) |
| 17:04 |
Taoki |
No stability in either the engine, nor the dev team. Hard to tell where the most bugs and crashes exist between the two :P |
| 17:05 |
hmmmm |
we need more manpower really |
| 17:05 |
Calinou |
sorry, not going to learn C++ anytime soon 8) |
| 17:05 |
|
Dragonop_ joined #minetest-dev |
| 17:05 |
Taoki |
That too |
| 17:05 |
Calinou |
my studies won't ever focus on it |
| 17:05 |
Krock |
Calinou, minetest runs on P2 (1997-2001) |
| 17:05 |
hmmmm |
as far as it stands right now we have two people basically running the show and then dozens of other people making small patches on what /they/ want (not what the project needs) |
| 17:06 |
Taoki |
MT has quite a few devs already, but it could use even more. Also, considering all the conflicts, more devleopers (meaning more points of view) might be even harder to manage than the current situation... |
| 17:06 |
hmmmm |
we don't just need more devs, we need high quality devs too |
| 17:06 |
hmmmm |
the problem with that is high quality devs usually have real life jobs and they're not able to dedicate too much time |
| 17:07 |
Taoki |
From a perspective of code knowledge, the core devs I'm familiar with are pretty good, from what I can tell |
| 17:07 |
Calinou |
yeah, high quality devs would rather work where it's profitable |
| 17:07 |
Taoki |
The story of open-source :) |
| 17:07 |
Calinou |
they're maybe pretty good but they have bad workflow or practices... maybe you can put up with that, but hmmmm can't |
| 17:11 |
hmmmm |
hmm |
| 17:11 |
hmmmm |
https://github.com/minetest/minetest/issues?q=is%3Aopen+is%3Aissue+label%3Ablocker |
| 17:11 |
hmmmm |
this is what we should be talking about |
| 17:12 |
|
BlockMen joined #minetest-dev |
| 17:12 |
hmmmm |
so est is probably working on then inventory bug so 3048 is accounted for, there's already a PR for 3042, and 3045/3064 probably need fixing as well |
| 17:13 |
hmmmm |
I am all for simple solutions but gregorycu's patch seems to just shift problems around instead of solving them |
| 17:13 |
nrzkt |
Calinou: OpenGL 4.2 is now supported by MESA fully, radeon and nouveau will support them soon |
| 17:13 |
Calinou |
right |
| 17:13 |
Calinou |
http://mesamatrix.net/ |
| 17:13 |
BlockMen |
hmmmm, #3045 is fixed by #3041 |
| 17:13 |
ShadowBot |
https://github.com/minetest/minetest/issues/3045 -- Glitching through ceiling |
| 17:13 |
ShadowBot |
https://github.com/minetest/minetest/issues/3041 -- Fix sneaking (fixes #329 and #665) by BlockMen |
| 17:14 |
hmmmm |
ahh that |
| 17:14 |
Taoki |
nrzkt: Not r600. Which is why I might be stuck with OpenGL 3.3 for god knows how long. |
| 17:14 |
hmmmm |
yeah that's a very heavy solution |
| 17:14 |
Taoki |
If you want an example of what NOT to do, High Fidelity is a great one in this domain (as much as I love the project): Making OpenGL 4.1 a requirement when not even the native Linux driver supports it. |
| 17:14 |
BlockMen |
node is just checked now if sneaknode changes |
| 17:15 |
hmmmm |
BlockMen, what about 3064? |
| 17:15 |
nrzkt |
Taoki r600 is not supported anymore by AMD as i remember |
| 17:15 |
Taoki |
Making OpenGL 2.1 a requirement though? Come on... of course that's a reasonable minimum version. |
| 17:15 |
hmmmm |
Taoki: I would be okay with making 2.1 a requirement |
| 17:15 |
Taoki |
nrzkt: Ah... good to know. The free driver isn't developed by AMD however, so this doesn't affect me. Linux devs offer support for much longer. |
| 17:15 |
Taoki |
Another good reason for me not to use fglrx. |
| 17:16 |
BlockMen |
hmmmm, you think its related? |
| 17:16 |
nrzkt |
Taoki radeon free driver is developped by AMD itself, not r600. AMD works on radeon and the new AMDGPU driver (which will land in Linux 4.2) for new cards |
| 17:16 |
Taoki |
hmmmm: If there are ever shaders that need that, or it would improve performance, sure. Obviously we don't need to make it a requirement just for the sake of it... only if it ever gets in the way of anything (my opinion) |
| 17:16 |
|
SopaXorzTaker joined #minetest-dev |
| 17:16 |
hmmmm |
BlockMen: Of course |
| 17:17 |
Taoki |
nrzkt: Ah, interesting. |
| 17:17 |
hmmmm |
well |
| 17:17 |
nrzkt |
Taoki: https://www.phoronix.com/scan.php?page=article&item=amd_bordeaux_strategy&num=1 |
| 17:17 |
hmmmm |
it has to do with sneaking, and this wasn't a problem before, all of a sudden it's a problem |
| 17:17 |
Taoki |
Well r600 hasn't been abandoned form what I read. And probably won't be |
| 17:18 |
Taoki |
It might of course be ported to the new drivers they're making |
| 17:18 |
BlockMen |
hmmmm, cant reproduce locally |
| 17:18 |
nrzkt |
AMDGPU is the new unified driver and this driver based upon mesa will unify Radeon on newer cards with FGLRX |
| 17:18 |
Taoki |
As in, I won't have a Linux update make my video card not work any more :P Cuz that would be crazy... especially since it's still rather new. |
| 17:18 |
Taoki |
Huh... isn't fglrx close-source? |
| 17:20 |
* Taoki |
will stick with the open drivers, whichever those are. |
| 17:20 |
hmmmm |
well then |
| 17:20 |
nrzkt |
yes but the new goal of AMD is to have a common part, i don't which part, but a common part. And Alex from AMD is the best Radeon contributor, we have OpenGL 4.1 on radeon because of him and 2 -3 other AMD developpers dedicated to opensource |
| 17:21 |
Taoki |
This is good to know... wasn't aware of some of that. |
| 17:21 |
nrzkt |
i use Radeon on archlinux and on Linux 4.1 + Mesa 10.6 it works very great, but stucked to opengl 3.0, we need mesa 11.0 to have opengl 4 |
| 17:21 |
Calinou |
FWIW, radeon requires a binary blob to deliver 3D and video acceleration :-) |
| 17:21 |
Taoki |
Yes |
| 17:22 |
nrzkt |
like in Intel and nvidia softs |
| 17:22 |
Calinou |
go boot a PC with an AMD card or IGP with Linux-libre and enjoy everything breaking |
| 17:22 |
Taoki |
Yeah, the firmware. |
| 17:22 |
Calinou |
nrzkt, this is the case of Intel Skylake too, now |
| 17:22 |
paramat |
VanessaE and all, how much need is there for pinenode aliases? (due to there recent renaming) pine nodes have only been placed by mapgen during 0.4.12dev, so perhaps there is no strict requirement for aliases? i would rather have none |
| 17:22 |
Calinou |
they added a required blob :/ and Maxwell will also require a signed blob with Nouveau |
| 17:22 |
Calinou |
(well, there's still zero Maxwell support in Nouveau for now) |
| 17:22 |
|
Zeitgeist_ joined #minetest-dev |
| 17:22 |
BlockMen |
hmmmm, cant reproduce on that server either |
| 17:24 |
hmmmm |
BlockMen, did you see the comments for #3041? |
| 17:24 |
ShadowBot |
https://github.com/minetest/minetest/issues/3041 -- Fix sneaking (fixes #329 and #665) by BlockMen |
| 17:24 |
Taoki |
I can live with the firmware being a blob. I mean, it's not like the physical graphics card is open-source either, so to say. It's mostly instructions on how to cross the wires, more so than usable software. |
| 17:25 |
Calinou |
the problem is that crippling (compute, overclocking, fan control) often happens in the firmware |
| 17:25 |
Calinou |
maybe you'd like to overclock your card, or remove crippling (eg. HDMI limitations)... |
| 17:25 |
Calinou |
it'd be much easier with free firmware, and wouldn't require binary hacks |
| 17:26 |
Calinou |
(have you seen binary hacks available for the proprietary NVIDIA driver? it's insane, but they exist) |
| 17:26 |
|
Zeitgeist_ joined #minetest-dev |
| 17:26 |
Calinou |
AMD has been caught crippling their HDMI adapter in fglrx |
| 17:26 |
Calinou |
but they didn't do so in radeon |
| 17:27 |
BlockMen |
hmmmm, how else should you get the highest point of a nodebox? each node can have multiple nodeboxes and guessing its size (like we did before) caused hovering over slabs, stiars, etc when sneaking |
| 17:27 |
Taoki |
Obviously it would be ideal. |
| 17:27 |
hmmmm |
BlockMen: ?? |
| 17:28 |
BlockMen |
you dont like the cb_max = 0, right? |
| 17:28 |
hmmmm |
huh no |
| 17:28 |
hmmmm |
that comment applied to all within the range |
| 17:29 |
hmmmm |
the act of calculating the maximum point in a nodebox is something that ought to be calculated once, not every single time |
| 17:29 |
hmmmm |
but I also added another comment saying it's fine that way for now since it's not executed every single step |
| 17:30 |
BlockMen |
sry, github didnt noticed me :\ |
| 17:30 |
hmmmm |
I'm just saying that I think ContentFeatures should have a new field like "highest_boundingbox_point" or something that's computed once, when the client receives the NodeDef |
| 17:31 |
hmmmm |
and that way you wouldn't have to recompute it each time |
| 17:32 |
BlockMen |
ah, yes. that makes sense. |
| 17:34 |
hmmmm |
BlockMen: another thing, in "if (!is_climbing) { ...} |
| 17:35 |
hmmmm |
you moved that part where speed in the Y direction gets bounded to 0, unconditionally |
| 17:35 |
hmmmm |
is that intentional? |
| 17:37 |
BlockMen |
yes |
| 17:38 |
hmmmm |
I'm not going to pretend I understand what's going on in that code :p |
| 17:38 |
hmmmm |
I hope this patch doesn't break anything else |
| 17:39 |
hmmmm |
well, maybe I do understand it. tell me if I'm wrong: |
| 17:40 |
hmmmm |
hmm no i don't get it |
| 17:40 |
hmmmm |
how is it possible for the player to go downward if sneaking |
| 17:42 |
BlockMen |
The collision detections seems to be broken at all (e.g. you are sinking in objects that have collide_with_objects) enabled. From what i found out this happens with sneaking aswell, so the speed = 0 is needed here to not glitch from the node when "beside" an sneaking |
| 17:43 |
BlockMen |
so it should be added "workaroud: remove when fixed collision" |
| 17:44 |
BlockMen |
*at sneaking only when "over" the edge of the node you are sneaking on |
| 17:44 |
BlockMen |
so we can either use this partial fix for now or just revert the last sneaking fix that got merged recently |
| 17:45 |
BlockMen |
idc which way around |
| 17:48 |
hmmmm |
well |
| 17:48 |
hmmmm |
which is more work |
| 17:48 |
hmmmm |
and which are you confident about having less bugs in the end |
| 17:48 |
hmmmm |
however you want to approach it is fine by me |
| 17:53 |
BlockMen |
hmmmm, since you pushed that. has it fixed #1551 or not? |
| 17:53 |
ShadowBot |
https://github.com/minetest/minetest/issues/1551 -- disable_jump group doesn't work on nodeboxes |
| 17:54 |
hmmmm |
I do not know |
| 17:55 |
hmmmm |
people report it as fixing the 1551, I didn't test it on my own |
| 17:55 |
hmmmm |
when I wrote "can confirm, works for me too", I was mistaken that casmir's trampoline mod uses nodeboxes |
| 17:55 |
BlockMen |
well, then i guess reverting is the safe method to get nothing else broken |
| 17:56 |
hmmmm |
? |
| 17:56 |
hmmmm |
your patch modifies the exact same lines that 5c3c965 touches |
| 17:56 |
hmmmm |
take a look at what it actually does https://github.com/minetest/minetest/commit/5c3c965c1a6e1fb5957d441c2d04d8470b80d710 |
| 17:57 |
hmmmm |
instead of checking the half-height of the player, it checks the collision box close to the player's feet |
| 17:57 |
BlockMen |
my changes more. and i have tested many cases, but the old is tested more. furthermore idk if i may break disable_jumping (again) |
| 17:58 |
BlockMen |
i ment "those lines and more", not "my changes" :\ |
| 17:58 |
hmmmm |
well test it out! |
| 18:02 |
|
Calinou joined #minetest-dev |
| 18:04 |
hmmmm |
guys, please take a look: https://github.com/kwolekr/minetest/commit/393345e6f78a21b4f7284f942ddb81066fbd2077 |
| 18:05 |
nrzkt |
hmmmm: what does our patch ? |
| 18:06 |
hmmmm |
er.... |
| 18:06 |
hmmmm |
#3066 |
| 18:06 |
ShadowBot |
https://github.com/minetest/minetest/issues/3066 -- SEnv: Remove static_exists from ActiveObjects in deleted blocks by kwolekr |
| 18:15 |
BlockMen |
hmmmm, why is static_block alsways 0,0,0? |
| 18:15 |
hmmmm |
it's that by default |
| 18:16 |
hmmmm |
if a value isn't specified |
| 18:17 |
|
Calinou_ joined #minetest-dev |
| 18:17 |
BlockMen |
but when this is supposed to be used only when deleting block from map, why having that parameter at all? |
| 18:17 |
hmmmm |
just for completeness |
| 18:18 |
hmmmm |
what if sometime in the future people decide to use it for something else? |
| 18:18 |
hmmmm |
shrug |
| 18:19 |
BlockMen |
then they add the param i guess? but beside that it seems good to me, but i have to say i have no clue how objects are handled serverside at all |
| 18:19 |
BlockMen |
or what happens when they are not set static |
| 18:20 |
BlockMen |
and disable_jump still works with my patch |
| 18:20 |
hmmmm |
see |
| 18:20 |
hmmmm |
it's difficult to say exactly what happens |
| 18:21 |
hmmmm |
ServerEnvironment is so intertwined with other pieces of minetest that it's practically impossible to understand the impact of one change |
| 18:21 |
hmmmm |
way too much state imho |
| 18:22 |
hmmmm |
BlockMen: I approve of your patch |
| 18:27 |
BlockMen |
ok, then i add the comment where speed is set to 0 and push it |
| 18:32 |
|
sloantothebone joined #minetest-dev |
| 18:38 |
BlockMen |
hmmmm, comment understandable? https://github.com/BlockMen/minetest/commit/7238df4c5952ed9eec491f424b8ca4b790cf7558#diff-efb51cde4bb3c805a95dc31b2b1781d0R197 |
| 18:41 |
hmmmm |
sure looks good. |
| 18:42 |
BlockMen |
thx. pushing now |
| 18:46 |
hmmmm |
https://github.com/kwolekr/minetest/commit/383e135ffed520e40393e25acf16762838ef7c9f PTAL |
| 18:50 |
BlockMen |
PTAL = please take a look? |
| 18:50 |
hmmmm |
yes |
| 18:50 |
BlockMen |
approved then |
| 18:51 |
hmmmm |
that was an original bug from 2012 :) |
| 18:51 |
hmmmm |
heh |
| 18:56 |
hmmmm |
alright so the rest is basically up to est |
| 18:56 |
hmmmm |
and then I think we can release |
| 18:56 |
hmmmm |
I'll try to figure out vanessae's playersao removal problem but no promises |
| 18:57 |
hmmmm |
the reason why I don't consider it a blocker is because it only affects vanessae and she keeps up to date with development versions |
| 18:57 |
VanessaE |
hmmmm: did you see that latest OOM log I paste(bine)d? |
| 18:57 |
hmmmm |
no, can you link me? |
| 18:57 |
VanessaE |
sec. |
| 18:58 |
VanessaE |
http://pastebin.com/KHETC1Wa |
| 18:58 |
VanessaE |
short version: on_playerReceiveFields() in the markers mod |
| 18:58 |
VanessaE |
891MB in use at the time |
| 18:58 |
hmmmm |
mmmm. |
| 18:59 |
hmmmm |
on_PlayerReceiveFields is a formspec thing, right? |
| 19:00 |
BlockMen |
yes |
| 19:00 |
VanessaE |
yeah |
| 19:00 |
VanessaE |
Sokomine says she pulls in a list of areas to fill some formspec field, but I know the areas database is less than 1 MB and under 5000 entries |
| 19:02 |
hmmmm |
891mb is such an odd amount of mem to crash on |
| 19:02 |
hmmmm |
but it could be a plausible amount for it to be the luajit limitation |
| 19:03 |
hmmmm |
the luajit limitation is caused by the range of the memory addresses allocated |
| 19:03 |
hmmmm |
so plausibly, just enough memory could've been allocated for non-lua purposes below the 2GB mark |
| 19:03 |
hmmmm |
and it didn't get a sufficiently low address from mmap |
| 19:04 |
hmmmm |
sokomine is female? |
| 19:04 |
VanessaE |
yeah |
| 19:04 |
hmmmm |
did not know that |
| 19:04 |
hmmmm |
well you know what they say about girls on irc right |
| 19:04 |
hmmmm |
(or more specifically, EFNet) |
| 19:05 |
VanessaE |
ha! |
| 19:05 |
hmmmm |
where men are men, women are men, and little girls are FBI |
| 19:05 |
hmmmm |
I bet sokomine is a fed |
| 19:05 |
VanessaE |
looks, btw, like jualit has a 1GB limit, not 2 |
| 19:06 |
hmmmm |
our leet hacking ring is going to get busted |
| 19:06 |
hmmmm |
VanessaE: it's disputed |
| 19:06 |
VanessaE |
and allegedly can be extended to 4GB with some compile option(s). |
| 19:06 |
hmmmm |
the official complaint says it's 2, I personally observe 2gb, and then some other people see it at 1 only |
| 19:06 |
hmmmm |
i think it heavily depends on the version of your kernel among other factors. |
| 19:07 |
hmmmm |
i think we may need to scrap luajit altogether at some point |
| 19:07 |
VanessaE |
also this, http://lua-users.org/lists/lua-l/2012-04/msg00729.html |
| 19:07 |
hmmmm |
the author retired and nobody is taking his place so I think it might become a dead project |
| 19:07 |
VanessaE |
oh that's bad |
| 19:08 |
hmmmm |
also somebody posted earlier a benchmark that shows that interpreted lua 5.3 is faster than luajit (!!!!) |
| 19:08 |
hmmmm |
will need to investigate |
| 19:08 |
VanessaE |
how's that even possible? |
| 19:08 |
hmmmm |
it's going to take a bit of integration effort though, since certain apis have been removed |
| 19:08 |
hmmmm |
I have no idea |
| 19:09 |
VanessaE |
"All of that said: don't worry, LuaJIT is in good shape and it's NOT |
| 19:09 |
VanessaE |
going away! It's open source, it has a great community. It just needs |
| 19:09 |
VanessaE |
a new management." |
| 19:09 |
VanessaE |
http://www.freelists.org/post/luajit/Looking-for-new-LuaJIT-maintainers |
| 19:09 |
VanessaE |
says CloudFlare has taken over "stewardship" of it) |
| 19:09 |
hmmmm |
it was being developed by a single person |
| 19:09 |
hmmmm |
nobody understands the code lol |
| 19:10 |
hmmmm |
anyway i'll have to check out the markers mod later |
| 19:10 |
VanessaE |
ok. |
| 19:11 |
VanessaE |
sokomine's gonna try to minimize the amount of items in whatever list it is that she's creating, but in the meantime the underlying cause still has to be fixed (if possible) |
| 19:12 |
hmmmm |
it's difficult to accidentally get into a runaway malloc situation with Lua |
| 19:12 |
hmmmm |
if anything I'm guessing markers does something odd to expose a bug in the core |
| 19:12 |
hmmmm |
man I dunno |
| 19:13 |
hmmmm |
I think I could do a whole lot more by submitting some patches to LuaJIT |
| 19:13 |
hmmmm |
backtraces on OOM are definitely possible but not without modifying their exception handling internals |
| 19:17 |
VanessaE |
inb4 <hmmmm> oh shit, I know why... this is stupid |
| 19:17 |
VanessaE |
;) |
| 19:17 |
hmmmm |
oh that's not gonna happen |
| 19:18 |
VanessaE |
sure.. spoil the joke :P |
| 19:26 |
paramat |
hmmmm or anyone about to merge to the engine? i would like to push 3061 |
| 19:26 |
hmmmm |
go ahead |
| 19:27 |
paramat |
thanks now pushing |
| 19:27 |
Sokomine |
hmmmm: the markers mod shows (when nothing else fits more cloesly) a list of all areas players protected on a server. this list can get long. vanessaes server has over 3000 areas. and my markers mod shows them all in a list in a formspec |
| 19:27 |
hmmmm |
WELL THEN |
| 19:28 |
VanessaE |
didn't I say that half an hour ago? :) |
| 19:28 |
Sokomine |
it's a bit surprising that such a long list works at all. i think i'll have to shorten the list for big servers so that it gets more convenient |
| 19:28 |
Sokomine |
oh, sorry :-) |
| 19:28 |
hmmmm |
try pagination? |
| 19:30 |
Sokomine |
well, apart from putting the string together for the formspec, there's also some collecting of which areas are to be shown inside a list and a sorting of said list using vector.distance and table.sort prior to composing the formspec |
| 19:31 |
paramat |
merge complete |
| 19:31 |
Sokomine |
hm, or just showing player names who protected areas. i think i'll have to find a more practical way to use that part (the normal use case is to show the current area and its extension) |
| 19:31 |
Sokomine |
and yes, i'm female. i've just learned long ago not to show that due to too many pre-puberty young boys getting too excited to see any female name :-/ |
| 19:40 |
Sokomine |
(on irc in general; not here) |
| 19:43 |
|
SopaXorzTaker joined #minetest-dev |
| 19:51 |
|
Amaz joined #minetest-dev |
| 19:58 |
VanessaE |
hmmmm: http://bayesanalytic.com/access-extra-memory-from-lua-jit/ |
| 20:13 |
paramat |
will push game#622 very soon |
| 20:13 |
ShadowBot |
https://github.com/minetest/minetest_game/issues/622 -- Default: New grass textures, new footsteps overlay texture by paramat |
| 20:13 |
|
luizrpgluiz joined #minetest-dev |
| 20:18 |
paramat |
now merging game 622 |
| 20:18 |
luizrpgluiz |
the rubenward is very smart, now I can limit the size of the world now |
| 20:18 |
luizrpgluiz |
:D |
| 20:22 |
paramat |
merge complete, no more depressing grass! https://cloud.githubusercontent.com/assets/3686677/9239537/91ae1006-4154-11e5-94e6-1b42d44d35e8.png |
| 20:24 |
luizrpgluiz |
that strange, only works in the release candidate of 0.4.13 |
| 20:28 |
VanessaE |
paramat: can you make the leaf decay interval and chance public so I can access them from moretrees? |
| 20:30 |
paramat |
i guess not worth it because leafdecay will be moved to engine soon according to est31 |
| 20:30 |
VanessaE |
ok |
| 20:30 |
paramat |
my PR was unpopular |
| 20:30 |
VanessaE |
how about at least reading a group so I can set how fast the decay goes? |
| 20:30 |
VanessaE |
e.g. leafdecay = 2 would be half as fast as = 1. |
| 20:31 |
paramat |
same answer =) |
| 20:31 |
VanessaE |
D: |
| 20:31 |
luizrpgluiz |
^^ |
| 20:31 |
VanessaE |
well let's just see how bad it gets. |
| 20:31 |
* VanessaE |
switches moretrees over to using default leaf decay settings. |
| 20:31 |
paramat |
eek |
| 20:34 |
paramat |
is there much actual need for pine node aliases? pine nodes have only been placed by mapgen during 0.4.12dev, so strictly we can leave out the aliases, for tidyness, i would prefer none |
| 20:34 |
VanessaE |
hard to say but already people have been using them in some worlds |
| 20:35 |
paramat |
hm yeah |
| 20:36 |
|
sloantothebone joined #minetest-dev |
| 20:36 |
Amaz |
I've made fairly extensive use of them in some buildings on a couple of servers. |
| 20:36 |
paramat |
okay stairs/slabs too? |
| 20:37 |
paramat |
yes okay i'll add the stair/slab aliases too then |
| 20:37 |
Amaz |
Yeah |
| 20:40 |
hmmmm |
lol |
| 20:40 |
hmmmm |
leafdecay was handled by the engine as well as grass growing |
| 20:40 |
hmmmm |
there was some kind of strange push to put it in lua... |
| 20:42 |
paramat |
! |
| 20:45 |
|
twoelk joined #minetest-dev |
| 20:54 |
VanessaE |
paramat: no appreciable change in CPU when letting a giant sequoia decay using default code |
| 20:55 |
paramat |
surprising but good |
| 21:00 |
paramat |
the lua load would be heavy though, but perhaps no different to a forest of little default trees |
| 21:12 |
sloantothebone |
I wish minetest would support 2d games |
| 21:14 |
hmmmm |
lol |
| 21:14 |
hmmmm |
I suppose it is possible by restricting view to third person only at specific camera angles |
| 21:15 |
kahrl |
... or formspecs >:-P |
| 21:15 |
sloantothebone |
No I mean squares instead of voxels |
| 21:15 |
sloantothebone |
Like this: http://zanzlanz.com/MineBlocks |
| 21:17 |
hmmmm |
well, it seems like you're looking for a different engine altogether |
| 21:18 |
|
BlockMen left #minetest-dev |
| 21:19 |
sloantothebone |
Do you know of such an engine? |
| 21:20 |
hmmmm |
no, and it seems like this isn't a #minetest-dev discussion |
| 21:20 |
hmmmm |
please stay on topic hee |
| 21:20 |
hmmmm |
here* |
| 21:20 |
sloantothebone |
https://scratch.mit.edu/projects/11375751/ this is paper minecraft, I like this because it darkens nodes that are out of site |
| 21:20 |
|
ElectronLibre left #minetest-dev |
| 21:21 |
sloantothebone |
I thought this was the place to discuss features of the engine |
| 21:21 |
hmmmm |
it's for development |
| 21:21 |
hmmmm |
discussions are typically of a technical nature |
| 21:22 |
sloantothebone |
Ok, can you tell me if its possible? |
| 21:22 |
sloantothebone |
Alright I'll talk about it in #minetest |
| 21:22 |
hmmmm |
it's absolutely not possible... this is akin to saying that your train is really nice but you wish it were able to go underwater too |
| 21:23 |
hmmmm |
what is possible (and quite easy, in fact) is to restrict camera angles and make the world only 1 node in a certain dimension so that it can emulate a 2d view |
| 21:24 |
sloantothebone |
Well that would work |
| 21:25 |
sloantothebone |
could we make lighting 2 dimensional too? |
| 21:25 |
sloantothebone |
The controls would have to be modified |
| 21:25 |
paramat |
we wouldn't want to support 2D stuff, try modding it instead |
| 21:26 |
sloantothebone |
How do you redefine the controls and the way the player moves in lua?? |
| 21:27 |
paramat |
ask in the other channel =) |
| 21:27 |
sloantothebone |
Ok |
| 21:27 |
twoelk |
2D? I would rather think everything is possible but will somebody ever do it? Probably not ever and actually I wouldn't know why somebody should as that would be a totally different game alltogether |
| 21:27 |
VanessaE |
why would you *want* to? |
| 21:28 |
|
kaeza joined #minetest-dev |
| 21:35 |
|
luizrpgluiz left #minetest-dev |
| 21:36 |
|
paramat left #minetest-dev |
| 21:59 |
|
luizrpgluiz joined #minetest-dev |
| 22:04 |
|
Donillo joined #minetest-dev |
| 22:04 |
Donillo |
Hello. I've sent minetest download page to my friend, he told me about minetest.exe saying "the program can not start because zlibwapi.dll is missing from your computer. try reinstaling the program to fix the problem" |
| 22:04 |
Donillo |
what could be reason for this? |
| 22:27 |
RealBadAngel |
missing library.... |
| 22:32 |
Donillo |
uh, what he can do? |
| 22:32 |
Donillo |
could it be wrong version? |
| 22:32 |
Donillo |
I don't run windows to test that |
| 22:32 |
|
paramat joined #minetest-dev |
| 22:33 |
paramat |
game#635 VanessaE does this look okay? |
| 22:33 |
ShadowBot |
https://github.com/minetest/minetest_game/issues/635 -- Stairs: Add aliases needed due to new pine wood name by paramat |
| 22:36 |
VanessaE |
I think so |
| 22:52 |
sloantothebone |
+CD-Server |
| 22:52 |
sloantothebone |
<Segfault22> no way to copy from minetest chat because some devs thought that we shouldnt have that as a feature -- but FALLING THROUGH THE CORNERS OF BLOCKS is just fine and dandy |
| 22:55 |
VanessaE |
this is a known bug, in irrlicht apparently. |
| 22:55 |
VanessaE |
(the copy&paste) |
| 23:08 |
|
thatgraemeguy joined #minetest-dev |
| 23:10 |
hmmmm |
you know what |
| 23:11 |
hmmmm |
minetest modding is following the path of evolution for graphics cards |
| 23:11 |
hmmmm |
>have fully-DIY code modules that let you do anything with tiny but efficient interface |
| 23:12 |
hmmmm |
>too slow! oh no, we'll just speed up commonly used routines by doing it in the core with more APIs |
| 23:12 |
hmmmm |
>too limited! a gazillion parameters |
| 23:12 |
hmmmm |
then we get to the equivalent of shaders |
| 23:13 |
hmmmm |
natively compiled .dlls instead of Lua |
| 23:13 |
hmmmm |
right now though, we have fixed-function rendering :( |
| 23:14 |
VanessaE |
heh |
| 23:14 |
hmmmm |
maybe it would be prudent to create a C++ api instead of a lua one |
| 23:15 |
hmmmm |
and then if anybody wants the Lua API back, then there could be a C++ mod that executes lua |
| 23:15 |
VanessaE |
maybe "in addition to" |
| 23:15 |
VanessaE |
but definitely not exclusive. |
| 23:17 |
paramat |
hmmmm some observations on the dungeon ring mystery https://github.com/minetest/minetest/issues/2926#issuecomment-131467196 |
| 23:18 |
|
deltib_ joined #minetest-dev |
| 23:19 |
hmmmm |
i think this may have something to do with the "voxelmanipulator flags" |
| 23:20 |
paramat |
yeah possibly |
| 23:20 |
paramat |
i stared at the code and wondered that |
| 23:23 |
paramat |
the flags are the only thing i could not get my head around, everything else seemed fine |
| 23:23 |
hmmmm |
VMANIP_FLAG_DUNGEON_INSIDE |
| 23:24 |
Donillo |
apparently my friend have just something weird with whole system, now he says that another game asks for libwinptheard-1.dll |
| 23:25 |
|
thatgraemeguy joined #minetest-dev |
| 23:27 |
paramat |
perhaps i'll rework the flags and see what happens |
| 23:28 |
paramat |
now pushing game 635 |
| 23:32 |
hmmmm |
honestly the dungeongen can use a lot of rework |
| 23:33 |
hmmmm |
I think the code that exists right there was just celeron messing around with the concept - it's not good code, but it tries out interesting things |
| 23:33 |
hmmmm |
was never meant to be perfect |
| 23:33 |
paramat |
! i love it as it is |
| 23:33 |
ShadowBot |
paramat: Error: You must be registered to use this command. If you are already registered, you must either identify (using the identify command) or add a hostmask matching your current hostmask (using the "hostmask add" command). |
| 23:33 |
paramat |
heh |
| 23:33 |
hmmmm |
ho hum |
| 23:33 |
hmmmm |
caves and dungeons seem to follow the sameish algorithm |
| 23:33 |
hmmmm |
start at some point |
| 23:33 |
hmmmm |
choose a random direction |
| 23:34 |
hmmmm |
make a room/cave/thing, choose a random direction starting from there, repeat |
| 23:34 |
paramat |
i just want to remove the rings, too common, and leave the complete/protruding dungeons as an option |
| 23:34 |
hmmmm |
it's nice for what it is, I guess, we shouldn't try to change it anymore |
| 23:35 |
hmmmm |
I think next we should try to make a buildinggen |
| 23:35 |
paramat |
could always add a second dungeon type |
| 23:35 |
paramat |
yeah |
| 23:35 |
hmmmm |
which is similar to dungeongen except "packs" the rooms together |
| 23:35 |
hmmmm |
maybe start out with a single rectangular space and make walls instead of make rooms and tunnels |
| 23:36 |
paramat |
i love stuff like this ^_^ |
| 23:36 |
paramat |
merge complete |
| 23:37 |
hmmmm |
we need to try generating structures based on worley noise sometime |
| 23:37 |
paramat |
mtgame devs are considering simple houses / small villages as part of default mapgen |
| 23:38 |
* paramat |
looks up worley noise |
| 23:38 |
hmmmm |
you know what a voronoi diagram is right? |
| 23:40 |
paramat |
yeah |
| 23:40 |
paramat |
aha i see |
| 23:40 |
paramat |
to divide a space into rooms |
| 23:41 |
hmmmm |
exactly. |
| 23:41 |
hmmmm |
for the distance metric instead of using euclidian we could use a taxicab space instead |
| 23:41 |
hmmmm |
that would make the rooms square-ish |
| 23:41 |
hmmmm |
taxicab distance i mean |
| 23:47 |
|
Donillo left #minetest-dev |
| 23:56 |
VanessaE |
hmmmm, RealBadAngel... can something be added to spawn_tree() to force trunks to overwrite other stuff as they're "drawn"? |
| 23:56 |
VanessaE |
(so that multiple neighboring trees don't have their trunks cut by intervening leaves etc) |
| 23:58 |
paramat |
sounds like paramat job =/ |
| 23:58 |
VanessaE |
ok paramat then ;) |
| 23:59 |
VanessaE |
I'm talking about the lsystem call though. |
| 23:59 |
twoelk |
wouldn't that generate the same effect just the other way round? |
| 23:59 |
VanessaE |
twoelk: trunks should cut through leaves, not vice versa :) |
