Time  Nick        Message
00:21 troller     #895 updated
09:42 adama_      hello all
12:24 celeron55   about the performance tests yesterday
12:25 celeron55   http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2011052013%20-%2001:35:20%20PM.png http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2011052013%20-%2001:40:26%20PM.png
12:26 celeron55   the removal of the texture atlas is likely to be a huge cause in that; also the generation of nodebox meshes inside solid ground
12:27 celeron55   (in the first one neither of those kick in, because there are no nodeboxes and there is only a single texture)
12:28 celeron55   my guess is that the missing texture atlas is the clear winner of bottleneckery in that
12:30 celeron55   but the key is that the slowness is measured as those two multiplied together; not added
12:31 celeron55   or maybe not the key; anyway, it increases the effect
12:32 celeron55   current 0.4 is probably faster than 0.3 in drawing monotonous simple scenery, and really clogs up when it's special features are used a lot
12:38 celeron55   in any case, i believe implementing a working and efficient texture atlas would fix that; if that exists, then the nodeboxes really don't matter almost at all as they all would be bundled in a single GPU call
12:38 celeron55   or, well, all of them per each mapblock
12:40 celeron55   VanessaE: you can try that by changing all content to use a single texture (same texture file name); it it runs fast, then texture atlas is the key
14:54 troller     idea: dynamically increase-decrease view range to match 30 fps
15:34 celeron55   what?
15:34 celeron55   it already does exactly that
15:34 celeron55   and has done exactly that since 3 years ago
15:54 Exio4       sapier should try to remove functions and loops, because a lot of people bad-uses them
15:54 Exio4       same with exceptions, and with the memory management, and why not pointers!
16:42 celeron55   Exio4: let's implement an improved version of java - without functions and loops
16:43 kaeza       celeron55, do you still have that branch for message translation around?
16:43 Exio4       celeron55: neat idea!
16:44 celeron55   kaeza: message translation?
16:44 celeron55   the one that implements localization of item descriptions in the certain way?
16:45 kaeza       that one
16:47 celeron55   i don't want it to be used as-is because it wastefully sends texts in every language to every clients; it should be reimplemented so that a client sends it's preferred language to the server and the server sends only those
16:47 celeron55   anyway, it's here; the WIP commit is a beginning of reworking it to the latter: https://github.com/celeron55/minetest/commits/modtranslate
16:47 kaeza       ah gotcha, and thanks :)
17:15 VanessaE    celeron55:
17:15 VanessaE    [11-06 12:06] <OldCoder> Without the audio patch...
17:15 VanessaE    [11-06 12:06] <OldCoder> I cannot take screenshots
17:15 VanessaE    http://minetest.org/mtpatch-131106a.txt
17:15 VanessaE    he means the second part of that .txt
17:16 VanessaE    can you comment?
17:17 VanessaE    [11-06 12:16] <OldCoder> VanessaE, FWIW The 2nd patch simply disables warning messages that flood my screen
17:17 VanessaE    [11-06 12:16] <OldCoder> The messages are there all the time and mess up the game
17:18 VanessaE    why are openAL messages even making it to the chat?  why are ANY error messages dumped there?  Put them in the F10 console if they absolutely must be on-screen
17:18 VanessaE    (and not just c55, anyone who wants to comment, of course)
17:26 celeron55   uhm... i guess the feeding of error messages to the chat could be removed
17:29 celeron55   it's just a few lines that deal with chat_log_error_buf in game.cpp
17:30 celeron55   just adding a third boolean parameter to ChatBackend::addMessage to determine whether they go to the non-console buffer would work fine
17:31 celeron55   i think it's useful to have them in the console though (because noticing errors is kind of important)
17:31 VanessaE    by console you mean F10, or the controlling terminal?
17:31 celeron55   F10
17:32 VanessaE    ok, yeah putting error messages there is probably okay as long as they don't constantly flood the console
17:34 celeron55   well if there's a constant flood of errors, it's not OK in any case
18:10 PilzAdam    ShadowNinja, I dont like that you revert commits from me without asking me
18:11 ShadowNinja PilzAdam: You weren't around. And it was a pretty commonsense thing, as otherwise you can't punch in creative.
18:11 PilzAdam    you can, simply use a tool
18:11 ShadowNinja I did get approval though.
18:11 PilzAdam    or, set dig_immediate=3 in the nodedef
18:12 ShadowNinja But ou still can't punch with the hand.
18:12 PilzAdam    normal digging is just way more common than triggering a punch event
18:12 PilzAdam    so using a tool to trigger punches is better than the very slow digging
18:12 ShadowNinja And even if there is a way around it most players won't find the way unless you thouroughly document it.
18:13 PilzAdam    nodes that really expect to be punchable can set dig_immediate=3
18:13 VanessaE    um, that's stupid, PilzAdam
18:13 PilzAdam    VanessaE, how so?
18:13 * Jordach   doesnt like instant digging
18:13 VanessaE    if a node has a punchable callback registered to it, it should automatically be excluded from any kind of dig immediate.
18:13 ShadowNinja That doesn't work with things like //p set and //mtschemprob start.
18:14 VanessaE    (i.e. either on_punch or register_on_punchnode or whatever the exact spellings were)
18:14 PilzAdam    the main goal of creative is to build and dig fast
18:15 ShadowNinja If you need to dig a whole lot of nodes fast you use //p set //set air.
18:15 PilzAdam    digging times of 0.5 are way too slow
18:15 PilzAdam    even some tools in non-creative mode are faster
18:15 PilzAdam    ShadowNinja, I do not use any external mods when I run minetest_game
18:15 VanessaE    what good is it to build stuff that you can't then operate?
18:15 ShadowNinja Yes, but it is understood that you can't punch with super-fast tools, and the solution is objious.
18:16 PilzAdam    well, you can operate it (as I said above)
18:16 ShadowNinja obvious*
18:16 PilzAdam    the digging is jut more common than interacting, so the digging should get the main focus
18:16 ShadowNinja So punching should be disabled?
18:16 PilzAdam    no, punching uses a routine that takes more time (i.e. take a tool)
18:17 ShadowNinja I am not the only one that didn't know that trick.
18:17 ShadowNinja You need to document it were players will see it.
18:17 ShadowNinja And where can ou document it?
18:17 ShadowNinja +
18:17 ShadowNinja +y
18:18 PilzAdam    in the mod description that depends on that feature
18:18 ShadowNinja (Bad 'y' key)
18:19 ShadowNinja So add "You need to use a slow tool to punch nodes" to every mod description? Not only that but using a differen't tool is kind of a workaround, not a solution.
18:19 PilzAdam    the 0.5 was choosen previously because of doors, they were opened with a leftclick; but since we have the on_rightclick() now this argument is obsolete
18:20 PilzAdam    you can also design the mod to work properly with the insta-dig in creative mode
18:21 ShadowNinja Although doors may not use on_punch a lot of other mods still do/
18:25 VanessaE    some nodes make no sense being operated with a right click
18:25 VanessaE    like TNT for example.
18:25 VanessaE    punch it with a torch.
18:26 VanessaE    and operating nods on right click has its own troubles
18:26 VanessaE    like having to code around the *wielded* node's on-right-click function
18:33 zat         ShadowNinja: I think I am done, wanna try?
18:45 * VanessaE  wonders what zat is done *with*
18:48 * zat       wonders too
18:48 zat         it is the SQL rollback
18:49 VanessaE    oh yes
18:49 ShadowNinja zat: Sure, can you put it on GitHub?
18:49 zat         is there a way that don’t require me to sign up on Github?
18:49 zat         really... I dont want to
18:50 zat         I have my own git server
18:50 PilzAdam    of course you dont have to sign up on github
18:51 ShadowNinja zat: You don't have to, what is the URL.
18:51 zat         hold on I am pushing it
18:55 zat         ShadowNinja: last commit on the master branch at git://bros.ivn.cl/mbarrera/minetest.git
18:55 zat         Now it won’t delete the old rollback.txt file, it will just create an empty file that serves as a “flag”.
18:55 zat         rollback.txt.done
19:33 zat         ok I just realized I could change the stack column into two, node and quantity, columns.
19:33 zat         that would optimize file size and speed further.
19:37 ShadowNinja zat: You add whitesoace to an empty line in SQL_databaseCheck BTW.
19:37 ShadowNinja whitespace*
19:37 zat         oh my god!!!!!!!!!!!!!!!!!!!!!!!!
19:38 ShadowNinja zat: What? Is it that hard to fix?
19:38 zat         No, it is hard to be relevant.
19:38 zat         ;)
19:38 ShadowNinja Also TXT_filecheck in the MODIFY_INVECTOTY_STACK section.
19:39 ShadowNinja And getEntriesSince. git diff isn't hilighting anything else in red...
19:40 zat         ShadowNinja: no need to tell me about those, they can easily be removed with a single regular expression.
19:40 ShadowNinja zat: Then use that regex. :-)
19:41 zat         Sure, next time I commit a change.
19:41 zat         I dont want to sound rude... but do you have any actually useful comment about the implementation?
19:42 zat         I need to know how it works, performs, on other systems.
19:42 ShadowNinja I like that the conversion speed display is more precise now, but I have only started testing it.
19:42 zat         well yes, it commits once every second.
19:45 ShadowNinja zat: 14:44:30: ERROR[main]: ServerError: LuaError: error: C++ exception -- After running /rollback_check 10 100000000
19:45 zat         let me try it
19:46 ShadowNinja It hapened after it was running for a while, trying to reproduce it...
19:47 zat         wont it happen again?
19:47 ShadowNinja I'm waiting on a second run...
19:48 ShadowNinja Yep, I can reproduce it. Just use a really big number for the time to go back.
19:48 zat         it must be sqlite refusing to accept such a big number
19:49 zat         I have not made the queries fault tolerant yet
19:49 ShadowNinja It should accept any 64-bit number.
19:49 zat         should!
19:50 ShadowNinja 100000000 is enough to trip it.
19:50 zat         we will have to see what itis
19:52 zat         ShadowNinja: now that I recall, for rollback_check the original code brings all the actions to memory and check their coordinates game-side.
19:52 zat         I left that part untouched...
19:53 zat         Probably the coordinates checking could be done SQL-query side.
20:09 ShadowNinja The rollback_check function returns a list of matching actions. Perhaps it should have a limit, eg the ten latest matches. And zat: /rollback_ckeck should definitely print the last 5-10 actions rather than only the most recent one.
20:18 zat         I will see if I can make it compare the coordinates query-side.
20:18 zat         because currently it *must* bring all matches and compare
20:50 sapier      https://github.com/minetest/minetest/pull/884 any coredev to review and judge this pull request?
20:58 sfan5       sapier: seems fine
20:59 sfan5       as long as there are no mistakes in the code :P
21:00 sapier      there are mistakes for sure ... first thing is I've written description.png in docs instead of description.txt ;-)
21:17 troller     maybe add #895 to 0.4.8 milestone?
21:19 ShadowNinja troller: 21 commits? You shouldn't be submiting pull requests in this state.
21:19 sapier      I'm not sure about it it changes lots of things if we're gonna release 0.4.8 soo this has to wait for 0.4.9
21:20 troller     ShadowNinja, 2 manth ago it was only one commit
21:20 troller     after 2-4 rebases again 1
21:20 sapier      you should squash it it won't be merged as is for sure ;-)
21:20 troller     i always squash before merge
21:20 ShadowNinja troller: Just use git commit --amend to begin with.
21:21 troller     it works with pushed?
21:22 troller     and look here https://github.com/minetest/minetest/pull/895/files
21:22 sapier      but that's a quite unimportant detail actually I'd be glad with performance improvements but I can't decide if there aren't any hard to see bugs hidden within it
21:23 sapier      and without some time of testing I'd not have a good feeling to add this to a stable release
21:23 troller     better to do nothing...
21:23 troller     100% bug free
21:23 sapier      it's not about bug free it's about stability
21:24 troller     my server works fine.
21:24 sapier      everything works fine within the environment it was created
21:24 troller     it was created on by notebook
21:24 troller     my
21:25 sapier      that doesn't make a big difference unless you run it on 6 servers using different os and versions ... do you?
21:25 ShadowNinja bind_address rebased.
21:26 troller     sapier, you check in this way every your commit?
21:26 troller     its game with 100 players
21:26 troller     and rules as in atomic station software
21:27 sapier      troller my comments are always based on the asumption everything to be added prior 0.4.8 is related to either fix critical bugs or beeing without any risk
21:27 troller     unable to play more than 5-10 players - critical and shame
21:27 sapier      was 0.4.7 better?
21:28 sapier      if not I'd not consider it "critical"
21:28 sapier      but of course that's my personal opinion I guess other core devs may disagree
21:28 troller     you never play in game too?
21:28 sapier      I do yet not on servers true
21:30 sapier      if you are confident enough these changes will not break things on windows or other os than yours I'm fine with merging but be sure I will remind you if it's not this way ;-)
21:30 ShadowNinja We aren't in a feature freeze yet, perhaps that should start on saturday or sunday and last a week unless more is needed?
21:31 sapier      a week?
21:31 ShadowNinja And I consider the most critical current bug the fact that sapling trees spawn without leaves.
21:31 troller     sapier, it os-independed
21:31 sapier      we have that many changes in there noone even tries to fix we need a month to fix them
21:32 ShadowNinja sapier: Two weeks maybe, but we can't wait a whole month.
21:32 sapier      have a look at my last bugfix ... the bug is os independent too yet it didn't occur on linux ;-)
21:32 sapier      if you guys help me fix the bugs 2 weeks might be enough but the last weeks I more feel ppl keep on adding new features (bugs) instead of fixing issues
21:33 init        ShadowNinja: we *need* to wait
21:33 init        some things need to get fixed
21:33 sapier      init what do you think needs fixing?
21:34 ShadowNinja init: Yes, but I think two and a half weeks is enough.
21:34 sapier      lol shadow are you dreaming? ;-)
21:35 ShadowNinja Well, whatever, but when should the feature-freeze start?
21:35 sapier      we haven't even decided how to fix the modstore hang issue and both possible options are far from minimal changes
21:36 sapier      imho I'd start it immediatly
21:36 init        that was i meant by "some things" sapier
21:37 ShadowNinja Both should be merged, but they need some modification to work together properly.
21:38 sapier      of course but imho kahrls changes are way more invasive then the async things
21:39 sapier      async only changes behavior of those calls related to the hangs while httpfetch changes almost any download
21:39 sapier      and async is limited to mainmenu (for now)
21:41 sapier      any other toughts to pro/cons of those changes?
21:43 ShadowNinja http_fetch should be available to Lua, asyncronously and syncronously, plus async Lua. http_fetch improves Minetest's downloading in more places than just the main menu.
21:43 sapier      thats crap
21:43 ShadowNinja ?
21:43 sapier      it's adding multiple async ways to api
21:44 ShadowNinja sapier: Is async Lua heavy at all?
21:45 sapier      yes it's heavy compared to calling a single function but lightweight compared to a download
21:45 sapier      so if you just want to multiply two numbers async is quite useless and only adds overhead
21:45 sapier      but overhead is minor compared to overhead added by download
21:47 ShadowNinja Well then only add http_fetch, and not http_fetch_async. And use async Lua.
21:48 sapier      that's what I intended to do
21:49 sapier      as kahrl is busy I'm guess I need to do that
21:49 ShadowNinja But the http_fetch branch must be merged for it's other improvements.
21:50 sapier      of course but without the mainmenu hangs it's not a blocker for 0.4.8
21:52 ShadowNinja sapier: Rearange the parameters to do_async so that multiple parameters or none can be passed. And does it suport multiple return values? Eg " return 5, {x=3, =4}, 'a string' "
21:52 ShadowNinja (Like minetest.after works)
21:52 sapier      no
21:53 sapier      that'd add useless complexity as I'd need to serialize a variable number of parameters passing a variable number and deserializing them
21:54 sapier      you can always enclose your parameters withing a single table ... you could even write a wrapper funtion in lua
22:24 celeron55   i think the fast digging was stupid; it was hard to dig single nodes
22:24 sapier      as far as I know it is already changed back?
22:25 celeron55   yes and PilzAdam complained
22:25 celeron55   read the log
22:26 sapier      :-) I guess I shoud read more carefully
22:33 sapier      any second agreement to 884?
22:34 sapier      PilzAdam if I remember correctly you've been one of the ppl complaining about unintuitive modmgr?
22:41 ShadowNinja sapier: The new layout looks much better
22:42 sapier      actually adding mod screenshot and description was intended at begining
22:42 sapier      there was a todo in code ;-)
22:43 ShadowNinja Yes, I figured that from the title.
22:43 sapier      hope you had a look at code too ;-)
22:43 BlockMen    hey, could someone agree that i push that? https://github.com/BlockMen/minetest_game/commit/4f474d9b29731f586e1fcbf26df55ece91c4c4ec
22:43 ShadowNinja Nope, just the screenshot. *looks*
22:44 sapier      argh ...
22:44 sapier      :-)
22:44 ShadowNinja BlockMen: No, that code if broken and ugly, the Lua treegen code has to be rewritten and fixed.
22:45 BlockMen    it has? why isnt it pushed?
22:45 ShadowNinja sapier: function tbl.splittext(text,charlimit) <-- What does this do?
22:46 sapier      it splits a long textline to a array of short lines (at least if I remember correct ... some time ago)
22:47 ShadowNinja sapier: Use string:split() then.
22:47 ShadowNinja sapier: Does no_screenshot.png already exist?
22:47 sapier      yes it's used in modstore
22:47 BlockMen    ShadowNinja, where is the rewritten code?
22:47 sapier      split doesn't work as required for this function
22:48 ShadowNinja Oh, I see what you mean.
22:48 sapier      argh there's debug code in there
22:48 sapier      that split function preserves newlines
22:48 ShadowNinja BlockMen: There is none. But the current code has to be redone. Use a schematic or read up on the LVM.
22:49 sapier      preserve means a newline will create a new line in array too
22:49 BlockMen    sapier, when testing #977 sometimes mt crashed without any error, warning, etc
22:49 ShadowNinja fgettext("no mod description available") <-- +Capital first letter and period.
22:49 BlockMen    so i cant tell you any reason :\
22:50 sapier      any idea what you did when it crashed?
22:50 BlockMen    ShadowNinja, why got this code pushed then? and how long will it need until someone rewrites it?
22:50 BlockMen    IMO it should be fixed, because trees dont grow now at all
22:51 ShadowNinja BlockMen: Whoever aproved it didn't read the code or didn't understand it. And that depends, can you fix it?
22:51 sapier      don't grow? they grew before so this needs to be fixed for 0.4.8
22:51 BlockMen    sapier, when opening modstore
22:52 BlockMen    https://github.com/BlockMen/minetest_game/commit/4f474d9b29731f586e1fcbf26df55ece91c4c4ec that fixes the sapling problem
22:52 ShadowNinja splittext can probably be rewritten to be nicer. But otherwise the code looks ok.
22:52 ShadowNinja They grow, but only trunks.
22:53 BlockMen    and fixes the instant grow of jungle sapling
22:53 sapier      I expected this wouldn't be very helpfull blockmen :-)
22:53 BlockMen    trunks != trees :P
22:53 sapier      would you mind trying to run a debug build within gdb?
22:54 BlockMen    gdb?
22:54 ShadowNinja sapier: Who? Of what?
22:54 sapier      talking to blockmen atm ;-) do you use linux or windows?
22:54 BlockMen    windows (only)
22:54 sapier      any other bugs in there shadow?
22:55 sapier      ok I'll test on windows have done most of testing in linux ... what build variant do you use blockmenß
22:55 ShadowNinja sapier: Looks OK, although I haven't run it.
22:56 BlockMen    release because debug doesnt work at all under msvc2010
22:56 sapier      hmm difficult ... I'm gonna try if this happens in 2012 too
22:57 sapier      yet my 2012 build is way more clean than official buildbot variant
22:58 BlockMen    you may test it a few times, most time nothing happens
22:58 sapier      puuhh .... that was close ... seems I canceled my accidental branch push to master soon enough
22:59 sapier      ok I do a linux test first maybe it happens there too but more rare
23:00 BlockMen    ok, back to my beginning question: is there a good reason to not fix tree growing?
23:00 sapier      as I said if it did work in 0.4.7 and doesn't work now we need to fix it ... it has to work in 0.4.8 at least as good as before
23:02 BlockMen    sounds like an agreement
23:03 ShadowNinja BlockMen: Your commit looks like an improvement, but the code needs a total rewrite.
23:03 sapier      I haven't had a look at your fix if it really does this (not things like e.g. adding new sorts of trees)... but if it does I'm fine
23:04 sapier      do I understand correct that most of those changes are moving from node.lua to tree.lua?
23:05 BlockMen    sapier, yes, because adding all that code in node.lua made it unreadable
23:05 ShadowNinja Yes, makes it hard to tell what has changed.
23:06 sapier      I'm almost everytime for spliting files (at least if it's not splitting a 10 lines file to two 5 line files)
23:06 BlockMen    so this commit seperates the growing code, moves the abms to functions.lua, fixes the growth rate of jungle saplings and tree grows
23:06 BlockMen    s/s/ing
23:07 BlockMen    imo the dirt and grass abm has to be moved to functions to, but i wanted not mix uo commits
23:07 BlockMen    s/o/p
23:08 sapier      can you tell what lines actually contain changes? it's difficult to find them ;-)
23:09 BlockMen    https://github.com/BlockMen/minetest_game/commit/4f474d9b29731f586e1fcbf26df55ece91c4c4ec#diff-fe24fdf96c7af2afc4edbda97f408e43R155 + 156
23:10 BlockMen    and the contentid is read correct now because it was not registered in curent code yet
23:10 BlockMen    ^ what caused the growing issue
23:11 sapier      ok I'm fine with it even if this may still need rewrite it's better than current code
23:11 BlockMen    k, i will add the new line at tree.lua and then push it
23:12 sapier      I assume noone is working at that better rewrite atm
23:13 ShadowNinja BlockMen: Um, where is the old ABM removed?
23:14 BlockMen    since nore wrote the code i checked his repo, but there was nothing new
23:14 BlockMen    ShadowNinja, nodes.lua
23:14 ShadowNinja Oh, there it is.
23:15 ShadowNinja BlockMen: Do you actually fix the treespawning code though?
23:16 BlockMen    yes, like i said already the problem was that the contentid was read before the node was registered.
23:17 BlockMen    so i fixed the junglesapling abm and restructured the whole thing, see above
23:17 BlockMen    i gonna push now
23:18 ShadowNinja Ok.
23:25 BlockMen    sapier, hmm...im not able to reproduce the crash at #977 atm
23:26 sapier      can you plz check the fix I just pushed it fixes a silly bug downloading any screenshot numerous times
23:27 BlockMen    sure
23:27 sapier      yet as this reduces complexity drasticaly the bug may  occur even less frequent
23:28 sapier      shadow did you manage to add error handlers everywhere?
23:28 ShadowNinja sapier: Not everywhere, I need help with script_run_callbacks.
23:29 ShadowNinja But almost everywhere.
23:29 sapier      yea that one is really tough
23:29 ShadowNinja It has been running on VE's servers for a few days and seems to work.
23:29 sapier      was hoping you could help with error handler in async calls guess I need to find out my self
23:30 ShadowNinja Except in the vector case -- The case it was written for.
23:30 sapier      funny
23:31 BlockMen    sapier, that fix is good
23:31 ShadowNinja It must be called from script_run_callbacks or similar. sapier: Check my 'pcall_errfunc' branch, but it has a lot of changes to facilitate the error handler, so my patch should be merged first.
23:32 sapier      async isn't changing a lot of things most cases it just adds things (except within mainmenu there unused variables are removed)
23:33 sapier      BlockMan yes but the original error may still be there just not beeing recognizable that often
23:36 sapier      I hope libcurl is threadsafe ;-)
23:46 BlockMen    well, lets see if i can reproduce anyway ^^
23:47 BlockMen    gd n8t
23:47 sapier      gd n8t