Time  Nick         Message
01:19 VanessaE     I'd like to remind those in channel of PilzAdam's/Taoki's VBO code ( https://github.com/PilzAdam/minetest/commit/7af479cba466f801c5cb648c8a449877957cb723.patch )  along with Exio4's "hardcode the value of 'e'" patch ( https://github.com/EXio4/minetest/commit/136d390905acaf6fb32f54b9583207f2a493ba4f )
01:19 VanessaE     together, these give me at least a 20 fps boost at the spawn on my server
01:20 Exio4        the hardcoding should be done when rba does 'its changes', right now it is useless and just 'bothers'
01:20 PilzAdam     I already told RBA to do that
01:23 VanessaE     just pointing it out.  the improvement is too good not to have.
01:24 Exio4        ah :P
01:29 NakedFury    I dreamed of minetest
01:29 NakedFury    was weird
01:30 NakedFury    it was evolved
01:30 NakedFury    and fixed you guys were in love with the source code rewrite and cleanliness
01:30 NakedFury    cant remember much
02:20 Exio4        small commit making bump-mapping compatible with 1.8, it is mainly for other thing that RBA will have to 'workaround' with his changes, https://github.com/EXio4/minetest/commit/c1d4467a132c1ee5673dcbc774421aca571d168e
12:04 BlockMen     celeron55, -> https://github.com/celeron55/minetest.net_dokutemplate/pull/9
17:12 sapier1      https://github.com/minetest/minetest/pull/640 can someone tell my why this isn't added?
17:13 sapier1      https://github.com/minetest/minetest/pull/774 https://github.com/minetest/minetest/pull/693 and https://github.com/minetest/minetest/pull/418 would be necessary to improove mobs too
17:14 PilzAdam     693 and 418 would be good for Simple Mobs too
17:15 sapier1      of course I always try to make improvements as generic as possible
17:16 sapier1      mobf development is on hold due to those patches not beeing included ... it's useless to improve minor issues while the big blockers aren't fixable
17:16 PilzAdam     Im not sure about the animation thing
17:17 PilzAdam     there is documentation in lua-api.txt missing
17:17 sapier1      it's most usefull combined with acceleration but you wouldn't have to calculate animation speed in simple mobs too
17:17 PilzAdam     and the fact that its not backwards compatible is really bad
17:17 PilzAdam     _really bad_
17:17 sapier1      it was backwards compatible but ppl complained about different functions ;-P
17:18 PilzAdam     I dont calculate the framerate in Simple Mobs at all
17:18 PilzAdam     I guess thats a mobf thing
17:18 sapier1      yes but you should ;-)
17:18 sapier1      no it isn't
17:18 PilzAdam     I have hardcoded speeds, thus I dont need to calculate it
17:18 sapier1      so your mobs have only fixed speeds full and stop?
17:19 PilzAdam     walking speed, running speed and standing, yes
17:19 sapier1      quite limited
17:20 PilzAdam     Simple
17:20 sapier1      ok mobf supports different speed levels so animation needs to be adjusted ... this can't be done by server as latency is way to big to get a reasonable result
17:20 PilzAdam     it looks good in Simple Mobs, I guess the dynamic speed is a not needed feature
17:20 sapier1      that's your oppinion
17:21 PilzAdam     and then there is this backwards incompatibility...
17:21 sapier1      I already had a backward compatible version that wasn't accepted ... you're driving me crazy guy ;-P
17:22 PilzAdam     why wasnt it accepted? and where is it?
17:22 sapier1      btw protocol will break anyway
17:22 sapier1      exactly same pull request before rewrite
17:22 sapier1      stepheight will do too
17:23 sapier1      and I don't see any other incompatibility ... maybe I'm wrong can you explain to me?
17:24 PilzAdam     I just see that you increase CLIENT- and SERVER_PROTOCOL_VERSION_MIN by 9
17:24 PilzAdam     and this change is not worth it
17:24 PilzAdam     why dont you simply add a new command, like GENERIC_CMD_SET_ANIMATION_WITH_PREDICTION?
17:25 sapier1      I don't have any objections ... but this will result in incompatibility too
17:25 PilzAdam     it at least doesnt break anything
17:26 sapier1      it breaks animation on all clients connection to a new server
17:26 PilzAdam     adding new things is always possible, chaning existing ones not
17:26 sapier1      it's still changed
17:26 PilzAdam     <sapier1> it breaks animation on all clients connection to a new server <- only if mods use the new function
17:26 sapier1      there's no new function
17:26 PilzAdam     of course you send the new function only if the new parameter to the Lua function is given
17:27 PilzAdam     s/function/command
17:28 sapier1      I don't think this is a good idea ... yes it keeps compatibility but adds inconsistency and lots of maintenance effort
17:29 PilzAdam     Ill merge the stepheight thing
17:29 sapier1      I don't have any chance to decide if parameter is intentionally nil or not ... so we need to add a new lua function just for adding a new parameter
17:29 PilzAdam     what would "intentional nil" do?
17:30 sapier1      what about 640? this is a major performance issue
17:31 sapier1      disable automatic update
17:31 PilzAdam     isnt that exactly when the old command would be send?
17:31 sapier1      most likely ... ok so this isn't an issue ... still sending two different commands for same thing is crap
17:32 PilzAdam     it isnt exactly the same thing, and its totally worth it to keep compatibility
17:32 sapier1      I'd even prefere to break animation completely on old clients
17:32 PilzAdam     " I'd even prefere to break" /me doesnt continue reading
17:32 sapier1      no it isn't ... you never tried to maintain a software containing years of compatibility fixes
17:33 sapier1      there's a point the only way to fix it is drop it completely
17:33 sapier1      I don't think many ppl use an outdated client and only those would recognize missing animation
17:34 sapier1      you should continue reading PilzAdam unless you want to drive away developery by arrogance
17:37 sapier1      so you're gonna keep cherrypicking what you think is needed and ignoring everything else ... fine ...
17:40 PilzAdam     stepheight=1 doesnt seem to work
17:41 proller      now impossible to connect 0/4/7 to git
17:41 sapier1      I didn't try for any height only 0.6
17:41 PilzAdam     I thought I could disable the jumping in Simple Mobs by setting stepheight to 1
17:42 sapier1      it's not meant to replace jumping ... now I understand why you've been willing to add it ...
17:43 sapier1      and I don't think it'd look good if it would work with that height
17:43 PilzAdam     oh wait, I made a mistake
17:45 PilzAdam     it works fine
17:46 sapier1      true just tried with 1.1 to ride ostriches
17:47 sapier1      what about get_surface?
17:47 sapier1      it's a small change but a huge performance gain
17:48 PilzAdam     stepheight pushed
17:48 sapier1      btw stepheight breaks protocol too
17:49 PilzAdam     as I said earlier, adding things is not breaking
17:49 PilzAdam     and I have added it to the bump to version 21
17:50 sapier1      you're wrong as soon as an additional member is added to entities this will break
17:51 sapier1      ok it's not gonna result in problems as long as long as no bogus data is ever removed ...
17:52 sapier1      guys this is plain wrong ... next to noone is using outdated clients with new servers .. and we're creating a maintenance nightmare to support it
17:53 PilzAdam     what happens if a new client connects to an old server?
17:53 sapier1      I guess nothing
17:53 PilzAdam     what does that mean? no animation?
17:54 sapier1      if done the way I suggest it'd change nothing if a new client connects to a old server
17:55 proller      push all breaking things now, before release
17:55 sapier1      at least if I didn't do a programming mistake but that could be fixed
17:59 sapier1      ok there's missing same try catch you rely on for your stepheight changes ... but that can be added quite easyly
17:59 PilzAdam     the get_surface pull seems good, but I recally hmmmm saying something about it
17:59 PilzAdam     cant remember what, though
18:00 hmmmm        it needs to be merged manually
18:00 hmmmm        there are conflicts apparently
18:02 sapier1      those changesets interfere
18:03 sapier1      if you want to merge more than one they need fixing
18:06 sapier1      I added documentation as well as a fix that enables new clients to connect to old servers without issues
18:08 sapier1      btw I don't even think old clients would have a problem as they'd just ignore the additional parameter in animation message
18:14 sapier1      https://github.com/minetest/minetest/pull/846 VanessaE could you check if this fixes your multiplayer tab issue?
18:21 celeron55    it should be made so that old clients will simply get no animation if it's used
18:21 celeron55    there's no good reason not to
18:22 sapier1      if I'm correct in reading that code old client's won't even notice
18:23 celeron55    eh
18:23 sapier1      to stop them from showing any animation we'd need to replace the old message by a new one and declare old cmd id deprecated ... that was what I meant with "Id even prefere .."
18:24 celeron55    how do you expect old clients to "not even notice" when you've made the server not even accept old clients?
18:24 celeron55    come on now
18:24 celeron55    stop the bullshit and make useful code
18:24 sapier1      I meant if the protocol versions wouldn't be changed the way ther#re changed right now
18:25 sapier1      I can't make usefull code if I don't even understand when a version change is required
18:26 celeron55    generally never
18:27 celeron55    anyway, quickly looking, yes, this seems such that it'll work on old clients too
18:27 celeron55    but i mean, the reason why breaking version cross-compatibility is discouraged is that most people don't ever use any single feature, so breaking some little thing doesn't generally affect anything
18:28 celeron55    people are much happier with a thing breaking that they never use than an old client not working because of something changed that they don't ever use
18:28 celeron55    8)
18:29 sapier1      I understand this but imho if keeping compatibility in some case means adding nonsense code breaking is better
18:30 celeron55    there would be need to improve the framework for easier maintenance of cross-version compatibility though
18:30 celeron55    i once messed around with a more robust serializer, called BKVL somewhere here: https://github.com/celeron55/minetest/commits/meta_set_nodedef_2
18:31 celeron55    maybe take a look for inspiration, but it's mostly a prototype
18:31 sapier1      of course this would be best ... but it's a quite a lot of work ... most likely breaking compatibility a (hopefully) last time again ;-)
18:31 sapier1      thx but I just finished formspec menu now I have to update mobf :-)
18:32 sapier1      ok almost finished
20:09 VanessaE     sapier1: no, that did not help.
20:09 VanessaE     (pull #846)
20:11 sapier1      hmm :-(
20:12 VanessaE     or at least, the IP that was already there got erased after I signed onto a singleplayer world, exited to the menu and went over to the "Client" tab
20:13 sapier1      no I guess this wouldn't help for that case ... I only fixed manually entered ips not beeing saved on connect
20:14 VanessaE     so the first run won't be saved, but subsequent ones will?
20:14 VanessaE     IP won't*
20:15 VanessaE     nope
20:15 VanessaE     doesn't save at all.
20:15 sapier1      no it wasn't saved before
20:15 VanessaE     start program:  type in an IP, go to single player, start a world, exit to menu, go back to client tab, IP is erased.
20:15 sapier1      it should be now ... unless something overwrites ... I'll have to do further investigations
20:15 sapier1      wait
20:16 VanessaE     that should be a pretty direct way to reproduce it
20:16 sapier1      it's not saved on switching tabs only if you really connect
20:16 VanessaE     it needs to be saved always.  never erase a field the user has typed something into
20:16 sapier1      even if it's invalid?
20:16 VanessaE     well I suppose if it's invalid :P
20:16 VanessaE     but I'd say "127.0.0.1" is quite valid :)
20:17 sapier1      I'll have to find out how to change this it's not meant to save anything on tab switching
20:19 VanessaE     even if I enter the IP and connect to it, it's still lost if I then go start a singleplayer world and come back
20:19 sapier1      hmm ok more than one error
21:20 proller      now old clients cant connect to new servers with nodebox version error, why to think about them?
21:44 sapier1      what does "World data version missmatch" mean?
21:44 BlockMen5233 PilzAdam, -> http://pastie.org/private/8ppvl70t8vbhljxqubdjow
21:47 sapier1      blockman you're right weather support breaks backward compatibility
21:48 Exio4        so, is weather the thing breaking my worlds?
21:48 PilzAdam     yes
21:48 PilzAdam     proller, you broke things
21:48 PilzAdam     not good
21:48 PilzAdam     BlockMen, its proller's stupid weather breaking the map
21:48 proller      its version up
21:49 proller      i have idea
21:49 sapier1      funny I'm accused for incompatibilities resulting in connection issues while we already broke world compatibility some days ago :-)
21:49 proller      for versions upping
21:49 Exio4        fix it then
21:49 PilzAdam     sapier1, s/we/proller
21:49 sapier1      it's been added so don't blame him
21:49 PilzAdam     ok, Ill blame celeron55 too
21:50 sapier1      at least if he didn't add this without asking anyone ;-)
21:50 PilzAdam     since he told proller to push it although 3 other core devs are against it
21:50 sapier1      obviously none realized the version sideeffect
21:51 proller      maybe make read in try-catch?
21:51 Exio4        you should read the version, and check it
21:51 PilzAdam     ^
21:51 proller      its imposible in old vesion
21:51 Exio4        if it is 1, or whatever it was, "read it in that way", if not, use the new
21:52 PilzAdam     Exio4, that doesnt help the old clients now
21:52 sapier1      if you increase number it won't work in old versions no way to add a workaround
21:53 proller      because now imbossible to increase version without broking
21:53 Exio4        are you serious?
21:53 sapier1      of course how should an old client correctly interpret new format?
21:53 Exio4        you added something, that could break compatability FOR MAPS?
21:53 proller      old can ignore new fields
21:54 sapier1      that's only half way
21:54 sapier1      this way you never ever can remove anything without causing crashes
21:55 proller      maybe support version+1 - for compatible changes, and make version+2 for big incompatible?
21:55 sapier1      only reasonable way is a new server doesn't enable higher version features on old maps thus those maps will work with old clients even after beeing used with new version
21:56 proller      and main thing: my changes dont save to disk, version was upped only for network
21:56 sapier1      and after two minor changes it becomes a big one? deciding +1 +2 is arbitrary
21:56 PilzAdam     proller, why is the new version written to disk then?
21:56 sapier1      is it serialization format?
21:56 PilzAdam     if you look at the files at the disk only, you increased the version without changing anything else
21:56 proller      because there was only one number for network and disk
21:57 PilzAdam     then split these numbers
21:57 proller      will try tomorrow
21:58 Exio4        hm
21:58 sapier1      I don't think this is gonna help
21:59 proller      maybe name next release 0.5.0 ?
22:00 proller      connection 0.4.7->0.4.8 broken, map broken
22:00 sapier1      wait
22:00 Exio4        ehm
22:00 sapier1      why did you increase the number at all?
22:01 proller      for transfer +2 variables for every mapblock
22:01 proller      only for network
22:01 proller      heat+humidity
22:01 sapier1      is the version detected from client?
22:02 sapier1      yes but if version isn't clients version I don't understand this version check
22:02 proller      seems yes
22:03 sapier1      hmm ok so version can be different things
22:05 sapier1      we should've put a version field to all our network messages