Time  Nick       Message
07:10 kahrl      #814: "It happens sometimes that the selection in a list follows the mouse (reproduceable by moving the scrollbar slider and then clicking outside of the formspec) (not in win build)"
07:10 kahrl      managed to reproduce this using a small irrlicht example http://paste.dy.fi/7nn
07:11 kahrl      so it has to do with the formspec disallowing focus change
07:13 kahrl      if you change canTakeFocus() to "return true;" the problem goes away
12:18 kahrl      PilzAdam: I figured out why the first problem in #814 happens: http://paste.dy.fi/kaj
12:18 kahrl      but I can't think of any way to fix it
13:36 PilzAdam   kahrl, its not a big problem, dont waste too much time on it
13:46 kahrl      I don't think I can even do anything about it unless we use a custom listbox
13:47 kahrl      ... which I'm planning right now (but for different reasons)
14:52 kahrl      ideas/design/examples for a "table" formspec element: http://paste.dy.fi/BJ9
14:55 nore       kahrl, I like it
15:17 celeron55  kahrl: tell us when you've used an equal amount for this that it would've taken to implement a client side lua based menu so we can all cry
15:17 celeron55  anyway yeah it looks fine for a formspec thing
15:18 kahrl      I don't think it will take that much work
15:18 kahrl      copying CGUIListBox gives a head start
15:18 celeron55  irrlicht has it's table element too; is it useless?
15:18 kahrl      I didn't look at it yet
15:19 kahrl      but I expect there to be complications :P
15:21 celeron55  sounds like a wise expectation 8)
15:21 kahrl      it can't be used as is because it doesn't support images, tooltips, or obviously indents
15:21 kahrl      but it might be better code to start from than CGUIListBox
15:23 kahrl      dang it, broke the irrlicht online docs
15:23 kahrl      Guru Meditation!
15:24 kahrl      IGUITable::getRowCount() "Get amount of rows in the tabcontrol." huh?
16:09 proller    https://github.com/proller/minetest/compare/heat
16:09 proller    https://github.com/proller/minetest/compare/liquid63
16:21 BlockMen   because im working on the formspecs atm i thought why not the menus too
16:21 BlockMen   what do you guys think of this -> http://i.imgur.com/3kD4HPM.png ?
16:22 ironzorg   german is hard
16:22 andersje   except for the fact that I can't read German, that's awesome.
16:22 rubenwardy nice
16:22 rubenwardy What about the chat menu?
16:23 BlockMen   not done yet
16:23 rubenwardy cool
16:23 BlockMen   first wanted hear if it is ok, else it would be work for nothing^^
16:23 PilzAdam   the buttons seem very randomly thrown at the screen
16:24 ironzorg   this ↑
16:24 ironzorg   it doesn't look like a menu
16:24 ironzorg   just a bunch of buttons
16:24 PilzAdam   also where are the default controls and debug info?
16:25 PilzAdam   the idea that it should cover the whole screen is good, though
16:25 BlockMen   kicked out, instead of that info the controls should be added to menu as setting
16:25 BlockMen   like in that pull request
16:26 BlockMen   ^IMO
16:32 thexyz     yeah of course we need controls and debug info displayed every time user presses [ESC]
16:53 nore       about menus I have a suggestion: a Lua function that declares a new button for the menu, that calls a custom callback
16:57 BlockMen   are the buttons better this way? -> http://i.imgur.com/D06Sw65.png
16:57 PilzAdam   yes
16:57 Calinou    "Pause Menu" is obnoxious imo
16:58 Calinou    on top of that, it's not even a pause
16:59 BlockMen   i guess it would be kinda har to make it pause for real (singleplayer)
16:59 BlockMen   *hard
17:05 celeron55  BlockMen: no it wouldn't (and that actually should be made)
17:05 celeron55  but for multiplayer it obviously shouldn't pause anything
17:08 BlockMen   celeron55, all the better. but i have no idea where to start
18:06 kahrl      what is m_Textures in GUIFormSpecMenu for?
18:07 kahrl      it's only ever pushed to but never read
18:07 kahrl      if I have to push the used textures to that list when implementing the table that'll make a lot of code ugly
18:21 kahrl      also: why are the texture names in image[], background[], image_button[], image_button_exit[] not escaped
18:54 sapier1    as far as I remember m_Textures is used to cleanup background textures on close of main menu
18:56 sapier1    I guess the cleanup is missing resulting in a texture leak
18:57 sapier1    are texture names escaped anywhere?
18:57 sapier1    maybe it's still a bug I cleant up much in formspec but not everything for sure ;-)
18:58 sapier1    kahrl if you do cleanup there could you add 4 to the textlist member cout ? docs allow skipping last two parameters but code doesn't
19:05 kahrl      hello sapier1
19:06 kahrl      I fixed the missing cleanup by moving formspec's texture management to ITextureSource
19:06 kahrl      ISimpleTextureSource to be precise, which implements only getTexture
19:06 sapier1    I guess then it's useless there
19:06 kahrl      yeah I removed it in formspec
19:07 kahrl      texture names are not escaped, would changing that break anything?
19:08 sapier1    I don't think there are any escape chars used there ... aren't "; , \ \n \r" dissalowed on most fs either?
19:09 kahrl      nope
19:09 kahrl      at least on linux I think everything but / is allowed
19:09 sapier1    yes but / isn't required to be escaped
19:09 kahrl      well sure
19:10 kahrl      but what if minetest is in a path that contains ;
19:10 sapier1    ok ok I'm convinced ;-)
19:10 sapier1    it should be escaped too
19:11 kahrl      not that other parts of the code are innocent in that respect, there are problems with loading the database when the minetest path contains non-ascii characters
19:12 kahrl      seems to be a problem in sqlite
19:12 sapier1    :-) yes the texture names are not unescaped ever since but that's no reason for not fixing it
19:14 sapier1    https://github.com/minetest/minetest/pull/846 any reason why this bugfix isn't merged yet? VanessaE seemd to be quite annoyed by that issue?
19:15 kahrl      what should I do about textlist? fix the docs or the code?
19:16 PilzAdam   sapier1, whats wrong with mainmenu.lua:505 and 506 in that diff?
19:16 sapier1    doc is wrong in any case as it doesn't mention 5 parameter textlists are allowed too
19:17 sapier1    I guess this is a whitespace fixup not visible correctly in github diff
19:17 PilzAdam   have you used tabs there?
19:18 sapier1    I don't have any idea its 20 days old ;-P
19:19 sapier1    as of raw view seems to be tabs yes
19:19 sapier1    8 spaces tabs shown for display argh
19:20 kahrl      8 is the unix default, I use it in vim
19:20 sapier1    whoever invented tabs should be punished to infinity for not specifying number of spaces
19:20 kahrl      just use spaces unless at the beginning of the line
19:21 PilzAdam   in Minetest a tab is 4 spaces
19:21 sapier1    8 is very very old default as it's almost useless for coding unless you use mixed mode indentions
19:21 kahrl      I have no problem with 8
19:21 kahrl      but I don't do insane nesting either
19:22 sapier1    ... 8 and 80 char limit is quite challanging ;-)
19:23 celeron55  anyone can use any kind of tabs they please, but line width limitation is calculated with 4 tabs
19:23 celeron55  and tabs should be used so that tab width doesn't affect formatting
19:24 celeron55  4-space tabs*
19:25 sapier1    I use tabs for "table style" initialization of multiple variables too (in lua) ... I know not quite consistent
19:25 celeron55  (in practice that means that if you want to align something in two lines, they should have the same tab indentation and spaces after that as needed
19:25 celeron55  )
19:29 sapier1    btw I need async lua events ;-) similar to this one https://github.com/sapier/minetest/tree/async_lua_functions
19:33 sapier1    reason is mobf adds a lot of on_mapgen actions done at once those add huge delays to map spawning ... but they don't need to be done instantly thus can be executed later
19:34 hmmmm      sapier1, can't you just add the information to some queue and execute them in server steps?
19:34 sapier1    atm I queue them in a global table and make global on_step handle those queued jobs until some time limit is passed ... of course this works but it'd be better if those could use multicore
19:34 sapier1    :-)
19:34 sapier1    already done
19:35 hmmmm      i've been thinking about multithreaded lua and i think we should go ahead with multiple interpreters
19:35 sapier1    if I remember correctly the async prototype creates a interpreter for each async job
19:36 hmmmm      for some projects in the past i had a dedicated async callback thread
19:36 sapier1    thus all information as well as the function to be executed needs to be serialized prior passing to that interpreter
19:36 hmmmm      yes, that isn't really as difficult as we initially made it out to be, though.
19:37 hmmmm      nor is it inconvenient
19:37 sapier1    have a look at the proove of concept implementation
19:37 hmmmm      besides, with client-side lua, we're going to have to develop a serialization thing anyway
19:37 sapier1    it's already capable of serializing functions
19:37 hmmmm      probably your thing is our best shot then
19:38 hmmmm      when i say we, i mean you, from the way it's looking i might not be able to work on minetest anymore with the new job unless i negotiate the contract
19:38 kahrl      how does it improve performance? thinking about e.g. envlock
19:38 sapier1    it needs some updating of course as it's quite old and  most important we need to sort out which calls can be used from async threads and which cant
19:38 hmmmm      karhl, computation bound things would benefit from multicore.
19:39 kahrl      ah well yeah
19:39 hmmmm      well
19:39 kahrl      I thought the spawning algorithms had to access the map
19:39 sapier1    but yes kahrl envlock limits benefit quite a lot
19:39 hmmmm      those could probably be done outside of an envlock
19:40 sapier1    yes kahrl ut maybe they could just read the part of the map they're interested in and work at that local copy ... not sure if this helps
19:40 sapier1    better option would be map access is locked only
19:41 kahrl      not sure how it works at the moment but aren't on_generated callbacks run from the EmergeThread?
19:41 kahrl      basically async
19:42 sapier1    if they are they still block main task due to envlock ... hmmm guess this means async threads will have same effect
19:42 hmmmm      he wants them to be asyncronous from the emerge thread though
19:42 hmmmm      erm
19:43 kahrl      they shouldn't have to lock envlock if all they do is use mapgen objects
19:43 sapier1    I want them to not interfere with map creation or mainloop
19:43 hmmmm      just FYI, i know you don't like unlocking the environment while in the script but just consider that it'd be safe to assume the data after you unlock doesn't instantly become invalid
19:43 hmmmm      it's sort of like map generating
19:44 sapier1    actually they result in huge delays ... but I can't tell for sure by what reason
19:44 hmmmm      you have no idea if the area you are currently generating is already there or modified or not, but you carry out the generation anyway and then 'merge' them together when you get back into the lock
19:44 hmmmm      and I really doubt the map is going to change in any significant enough way to modify the path or areas to place
19:44 hmmmm      within the time you take to computei t
19:44 hmmmm      compute it*
19:45 sapier1    and what I can tell too is the variant queuing the spawn algorithms and doing them in on_step loop doesn't result in this big delay (most likely because the main loop doesn't need to call all jobs in a single step)
19:47 sapier1    disadvantage is if your server crashes for some other reason while jobs are queued in lua they're lost forever
19:53 VanessaE   kahrl: new bug for you....
19:54 VanessaE   just noticed this today, might be caused by your diagonal flowing fix:  http://digitalaudioconcepts.com/vanessa/hobbies/minetest/screenshots/Screenshot%20-%2008202013%20-%2003:53:39%20PM.png
19:54 VanessaE   I don't think the water's edge is supposed to do that :D
19:54 sapier1    looks like a monster face ;-)
19:55 PilzAdam   VanessaE, do you use new_style_water?
19:55 VanessaE   yup
19:56 VanessaE   I do
19:56 PilzAdam   then dont use it
19:56 VanessaE   sapier1: haha
19:57 PilzAdam   this has nothing to do with the diagonal water
19:57 VanessaE   PilzAdam: I'm pretty sure this started with that commit or very recently at least.
19:57 VanessaE   it's never done that before that I remember seeing.
19:58 PilzAdam   I know it does it since I fixed new_style_water
19:58 VanessaE   hm, guess I never noticed it before then
19:59 kahrl      something is trying to load the texture "\0" in the main menu
19:59 kahrl      strange
19:59 sapier1    no it isn't
19:59 sapier1    ohhh
20:00 sapier1    ok it is ... I should read more precise after answering :)
20:01 kahrl      ok well here's why
20:01 sapier1    some empty variable?
20:02 kahrl      unescape_string is still not fixed (I found that bug months ago and forgot about it)
20:02 kahrl      "i <= s.length()"
20:02 sapier1    :-)
20:18 kahrl      committed what I have so far: https://github.com/kahrl/minetest/commit/ae81adbeef9593ca2e68175937c6991207e1f363
20:18 kahrl      ^ sapier1, in case you want to look
20:21 kahrl      working on the table thing and textlist now
20:21 sapier1    diff is terrible on that I can't even tell whats new and what got moved only
20:21 kahrl      hmm, dunno what to do about that
20:22 kahrl      basically TextDestGuiEngine and MenuMusicFetcher were moved, the other things are actual changes
20:22 kahrl      I moved them because they caused a bunch of headers to be pulled in
20:23 sapier1    ok maybe you should do this in different commits?
20:23 kahrl      maybe
20:23 sapier1    it's hard to review those changes if different things are mixed
20:25 kahrl      does git have a command "git add --lines X,Y,Z µ.cpp"?
20:25 sapier1    I don't know I'm still new to git
20:26 PilzAdam   kahrl, try --interactive
20:26 kahrl      PilzAdam, thanks
20:29 kahrl      now how do I use this thing
20:29 PilzAdam   I never used it
20:30 PilzAdam   git add --help, see "Interactive Mode"
20:31 kahrl      tl;dr I'll do this by hand
20:46 kahrl      sapier1: https://github.com/kahrl/minetest/commits/isimpletexturesource2
20:48 sapier1    way more easy to read yes :-)
21:30 proller    https://github.com/proller/minetest/compare/heat
21:30 proller    https://github.com/proller/minetest/compare/liquid63
21:30 proller    need to make bot for autoposting...