Time  Nick       Message
04:08 hmmmm      :/
04:08 hmmmm      so I'm testing PseudoRandom in Dieharder
04:08 hmmmm      so far it failed operm5, bitstream, opso, oqso, parking_lot, 2dsphere, 3dsphere, squeeze, craps
05:04 hmmmm      here are the full results:  http://fpaste.org/198380/64822551/
05:11 Tablet_One I think I found a bug in the newest android apk on github- when selecting a server from the server list, the address and port are not transferred to their respective input boxes, requiring the port and address to be typed manually
05:34 hmmmm      for comparison, here is the PCG32 rand algo:  http://fpaste.org/198382/42648275/  and here is FreeBSD's libc rand(): http://fpaste.org/198383/14264828/
05:36 Tablet_One Also, the game crashes upon selecting the inventory
05:43 hmmmm      hah... oh that's nice
05:44 hmmmm      great job, android maintainers another quality minetest release.  i'm sure it'll get an excellent rating in google play
05:54 Tablet_One Heh
08:55 nrzkt      VanessaE, this doesn't stop the death loop. This repair the player which are stucked with 0 hp when connects. The deathloop is not find yet. I think a trigger is missing somewhere, but i need to find where
08:55 VanessaE   nope.avi
08:55 VanessaE   doesn't fix them when they connect, either.
08:55 nrzkt      kahrl, sfan5 , ShadowNinja : #2495, agree ?
08:55 ShadowBot  https://github.com/minetest/minetest/issues/2495 -- Save creative_mode and enable_damage setting for each world in world.mt by fz72
08:56 VanessaE   wat
08:56 nrzkt      then it's another problem if this patch doesn't fix the loop
08:56 VanessaE   why?  that's specified in minetest.conf
08:56 VanessaE   wait, nevermind.
08:56 VanessaE   that's client-side/singleplayer
08:57 nrzkt      yes
08:57 VanessaE   anyway yeah, your fix doesn't actually work in practice :)
08:57 VanessaE   there's still only one way to break out of the loop - keep damage enabled on the affected server.
08:57 nrzkt      okay. I know the symptoms of the bug but somethink is missing somewhere on respawn
08:57 VanessaE   mind you, having damage disabled should have also prevented what started the loop in the first place (the set_hp() call)
08:58 VanessaE   but I patched around that bug in technic so at least *that* can't happen anymore
08:58 nrzkt      but the think you said is good, maybe there is a enable_damage somewhere which doesn't permit to respawn players
08:58 VanessaE   yes indeed, the player never actually re-spawned.
08:58 nrzkt      because they mustn't die and the code don't handle it
08:59 VanessaE   but the more pressing thing that needs fixed (and this isn't your department I guess) is this ghost entity + player object delete issue I've been bitching about
09:00 VanessaE   between that and the player death loop, well, you can understand how frustating it is.
09:00 nrzkt      in respawn we have this test
09:00 nrzkt      if (!player->isDead() || !g_settings->getBool("enable_damage"))
09:00 nrzkt      enable_damage will be removed
09:01 nrzkt      this doesn't permit to respawn player, because if player is dead and no damage the packet is ignored
09:01 nrzkt      i haven't seen it before, sorry :) i fix it now
09:02 nrzkt      pushed.
09:02 nrzkt      you can patch and remove your workaround :)
09:03 VanessaE   nope, can't
09:03 nrzkt      maybe, but later
09:03 nrzkt      why i don't have voice today ? :o
09:03 VanessaE   unless that ^^^^ idea also causes the engine to ignore the set_hp() call :)
09:03 VanessaE   idk
09:04 nrzkt      in fact, damage are handled somewhere, but we have a setHP call
09:04 nrzkt      and this ignore the enable_damage call
09:04 VanessaE   well I mean the lua end
09:04 nrzkt      i think having a set_damage call which handle enable_damage paramet would be better, no ?
09:04 VanessaE   nah.
09:05 nrzkt      this permit to remove the get_hp/set_hp calls and do directly the damage with a call which could disable itself damages on non damage servers :)
09:05 VanessaE   if you call someplayername:set_hp() to kill the player, and damage is not enabled on the server, well, the set_hp() command should be ignored and the player's HP should be reset to infinite
09:06 VanessaE   but instead, that set_hp() call actually sets the HP, even with damage disabled.
09:06 VanessaE   that's how I was able to trigger the loop
09:06 VanessaE   (in this case, technic corium, chernobylite, and other radioactive items)
09:07 nrzkt      i see
09:07 nrzkt      but in fact, setHP doesn't use enable_damage, because it's an HP setter, not a damage generator
09:07 VanessaE   in technic I've patched around it to check for the enable-damage flag and skip the radiation-injury ABM if damage isn't enabled
09:07 nrzkt      if we have a set_damage call for radioactive... etc it permit to disable the damages on disabled damages servers, directly
09:08 nrzkt      i think it's the job of core to do it, not mods, many mods will duplicate your workaround whereas we can offer a better path
09:08 VanessaE   agreed.
09:08 VanessaE   but new APi calls would be overkill
09:10 VanessaE   now, if there were a do_damage() which takes a relative damage amount, that would be preferable to the absolute value needed by set_hp(), but what's there in technic works well enough that that isn't needed right now
09:10 nrzkt      for technic, but other mods can use it :)
09:10 VanessaE   plus any such changes would need a damage system that allows for, say, floats for HP, or a higher-than-20 range
09:11 VanessaE   so that's all stuff for later imho
09:11 nrzkt      0.5 yes
09:11 VanessaE   (this subject has been covered before
09:11 VanessaE   )
09:11 nrzkt      i set it on the 0.5 roadmap :)
09:11 nrzkt      then i will propre a PR for doing damage later today
09:12 VanessaE   others have suggested a higher-than-20 range is better, with an open-ended max.  I lean heavily toward floats within the 0-20 range, for backward-compatibility sake.
09:13 nrzkt      i don't study the HP system, but i think there is something to do client side which break the compat with old clients
09:14 VanessaE   well, we'll see.
09:14 VanessaE   if it can be done without breakage, all the better
09:15 nrzkt      yes :)
09:20 VanessaE   bbl, bed
09:33 nrzkt      sfan5 ?
11:01 nrzkt      i will push a little code cleanup, moving some Server Send() code to functions instead of building packets into the ASyncRunStep
11:18 nrzkt      last google stats (saturday): we have 6 customers on Play Store. 4 Android 4.4 and 2 Android 2.3
11:27 Calinou    when do you get new stats, weekly?
11:28 nrzkt      Updated every day for today - 2 days
11:29 nrzkt      and this is a graph, then i can get 1 month or more stats
11:43 sfan5      nrzkt: ?
12:35 nrzkt      sorry sfan5 i was AFK. Can you look at the PR i mentioned to you on github ?
12:36 nrzkt      and if you can add the Android apk download link there: http://minetest.net/download
12:54 everamzah  get a segfault on wrong password now
12:55 nrzkt      with current master ?
12:56 nrzkt      no problem on my build
13:01 everamzah  yes current master and it says "Bus error"
13:01 nrzkt      what is your system ?
13:02 nrzkt      Bus error ? :o that's very strange
13:03 kilbith    can't reproduce it either
13:04 nrzkt      if you can use gdb and give us a backtrace
13:05 everamzah  it gave proper formspec "wrong password" first gdb run and then i entered the password wrong again and this time terminal says program receieved signal sigsegv -- where do i paste this?
13:05 nrzkt      now
13:06 nrzkt      type "bt f"
13:06 everamzah  ok
13:06 kilbith    everamzah, pastie.org
13:06 nrzkt      you will have the backtrace, and use pastie.org
13:06 everamzah  no symbol table info available, screen stuck on "connecting to server"
13:06 everamzah  oh
13:06 everamzah  ok thx
13:07 everamzah  http://pastie.org/10030288
13:10 nrzkt      segfault when you leave game ?
13:12 nrzkt      it seems it's a problem with your locale sqlite database
13:13 nrzkt      local*. A statement is empty
13:14 nrzkt      ShadowNinja will be more competent than me for this. If you can compile minetest in debug mode it's also more useful
13:14 everamzah  ok will do
13:23 everamzah  http://pastie.org/10030331 <-- the first time when in gdb it doesn't segfault but gives "wrong password" dialog, but the second wrong password is when it segfaults.  when not running in gdb it segfaults on the first wrong password
13:24 nrzkt      => FINALIZE_STATEMENT(m_stmt_begin)
13:26 nrzkt      it seems that your sqlite backend was not well initialized. verifyDatabase() will init it. I don't have more information on sqlite, i'm not specialist
13:36 kahrl      Looks like #2219 came back from its grave
13:36 ShadowBot  https://github.com/minetest/minetest/issues/2219 -- Singleplayer segfaults if server rejects client
13:36 kahrl      the reason is probably that not all statements are initialized to NULL by the Database_SQLite3 constructor
13:37 kahrl      (introduced by 708337dfc2)
13:41 kahrl      https://gist.github.com/kahrl/6055560b335017056248
13:43 kilbith    kahrl, btw : https://github.com/minetest/minetest/issues/2496
13:44 kahrl      kilbith: I'm a bit occupied at the moment, but I might do it in April if nobody has by then
13:46 kilbith    now that we have more space in the list, we could add the "signal" icon for the lag
13:46 kilbith    (if it's accurate enough)
13:49 Tablet_One No offence, but I think the Android build should've been tested more before being added to the Play store- #2497 is a pretty big problem
13:49 ShadowBot  https://github.com/minetest/minetest/issues/2497 -- Port and address must be manually entered on Android
13:51 nrzkt      i haven't seen this issue on my client, i retest to be sure
13:51 nrzkt      you are right.
14:02 srifqi     so, any chance to fix?
14:46 est31      I think we should talk about how android builds are released
14:46 est31      because currently its very manual
14:46 est31      nrzkt announces it on chat, and that is it.
14:47 est31      perhaps tagging each "release"?
14:52 nrzkt      srifqi, Tablet_One yes
14:54 nrzkt      est31, you are right, it's a little bit manual, we released it as soon it was technicaly possible, based on 0.4.12, but 0.4.12 is not really android Friendly
15:06 est31      so you say that this release is manual, and the releases after that are "safe" for android?
15:06 est31      about the nightly builds, where can I download them?
15:08 nrzkt      atm there are no upload. Jenkins build every night a build from master here: http://jenkins.unix-experience.fr/job/minetest-android-OfficialRepository-RollingRelease/
15:08 nrzkt      i need to add an upload task to a place, like my minetest-ftp.unix-experience.fr/android for example
15:09 sfan5      nrzkt: github gives me a list of all issues w/ new stuff since last time, this is why i don't notice when I'm mentioned
15:09 sfan5      nrzkt: also you need to mention me by "@sfan5", just "sfan5" is not enough
15:09 nrzkt      sorry then sfan5, but i use @sfan5 on my posts :)
15:09 sfan5      you didn't here: https://github.com/minetest/minetest/pull/958#issuecomment-77155144
15:10 nrzkt      right, but later i do it in this post :p
15:10 nrzkt      sorry for the mistake on first ping :)
15:11 nrzkt      est31: i will look at this when i get more time, i'm at work now, it a simple task to do on jenkins :)
15:12 est31      ok thanks nrzkt.
15:14 nrzkt      est31: i add the archive mode to jenkins, we will see this night where the file will be added. I also changed the rolling release mode to debug, else you can't install the build on your phone without sign the APK :)
15:36 sfan5      nrzkt: did you change something w/ LocalPlayer::move?
15:44 sfan5      this sucks
15:44 sfan5      why isn't GameRunData a member of the Game class
15:45 sfan5      it's passed to almost any function
15:45 sfan5      but nobody had the idea to just make it a member of the class
15:45 sfan5      now i need to pass it along with many more params just because my function needs it
15:46 sfan5      it's also not consistently named
15:46 sfan5      in function a it's called interact_args
15:46 sfan5      in function b runData
15:46 sfan5      in function c run_data
15:46 sfan5      or in function d args
15:46 sfan5      I'm not even making those up
15:48 sfan5      i have to add 3 lines just to get this damn variable into my function
15:48 nrzkt      sfan5: no i generally work on server, client is not my first priority
15:48 sfan5      </rant>
15:49 nrzkt      some designs are broken. PlayerSAO and Player must be merged to, i think . And we need to have a PlayerSession
15:50 sfan5      PlayerSAO is mostly older code
15:50 sfan5      but this GameRunData thing is new
15:50 sfan5      and still badly designed
15:51 nrzkt      it's not mine, i don't know this object
15:51 sfan5      i wasn't implying that it was your fault
15:51 nrzkt      but you can improve the object, why not improve it ? :)
15:52 est31      the selection bug is somewhere in guiTable.cpp I guess
15:52 nrzkt      for Android ?
15:53 est31      yes
15:53 nrzkt      i think it's in tab_simple_main.lua it miss a handler for click somewhere if i understand
15:53 est31      but I need more testing
15:53 sfan5      ~tell Zeno` http://irc.minetest.ru/minetest-dev/2015-03-16#i_4188723
15:53 ShadowBot  sfan5: O.K.
15:54 est31      nrzkt, there is an event handler, but its not called with "CHG" as its supposed to
15:58 est31      oh
15:59 est31      it might be something different too
15:59 est31      on desktop, the event is simply "CHG:2"
15:59 est31      on mobile, its "CHG:2:9"
15:59 est31      and then everything explodes
16:15 est31      somehow I believe the bug wasn't with android but with normal desktop version
16:15 est31      inside the API
16:15 est31      and then lua assumed to be broken
16:23 nrzkt      i don't know very well this API, but we need to fix this bug.
16:23 est31      I'm on it
16:23 kahrl      that's rather weird, because it's a table in both tab_simple_main and tab_multiplayer
16:23 est31      now I suspect lua again :)
16:23 kahrl      "CHG:2" would be something a textlist sends, not a table
16:24 est31      kahrl, look at core.explode_textlist_event
16:24 kahrl      yeah, I see that now
16:24 est31      aww
16:24 kahrl      tab_simple_main should call explode_table_event and not that
16:24 sfan5      nrzkt: i rebased #958
16:24 ShadowBot  https://github.com/minetest/minetest/issues/958 -- Add player list, viewable by holding Tab by sfan5
16:24 est31      yup
16:24 est31      thats the fix
16:25 est31      man 2 hours searching in the dark and then a one liner ...
16:25 nrzkt      sfan5: thanks for the rebase, but the question was more: do we need this fix in core or use Lua ? i agree this function must be in minetest, core or Lua i don't care :)
16:26 sfan5      nrzkt: which fix?
16:26 nrzkt      this PR*
16:26 sfan5      it would be easier to add it to core
16:26 sfan5      but using lua would be better
16:27 est31      so, I do a PR kahrl, or you fix it?
16:27 est31      (#2497 that is)
16:27 kahrl      I'm a bit busy, so go ahead and make a PR
16:27 kahrl      you worked longer on it than I, you get the attribution :)
16:28 nrzkt      ok est31 as soon as you send it and i'm at home (in 1-2 hours) i test the build on my phone and if it's okay for Android and regular clients i push it to stable-0.4 and bump a new apk
16:28 kahrl      (also I can't test android at all :P)
16:28 est31      dont forget to push it to master too
16:29 nrzkt      yes, i will do the two push
16:39 est31      wait thats not the fix
16:39 est31      my pr
16:39 est31      ill fix the fix pr :)
16:44 est31      fixed
18:34 nrzkt      ~tell est31 now testing your PR on my phone and android emulator
18:34 ShadowBot  nrzkt: O.K.
18:42 nrzkt      ~tell est31 your PR is okay and works perfect. I push it
18:42 ShadowBot  nrzkt: O.K.
18:47 nrzkt      i do a final apk test with release mode enabled in stable-0.4 and i push the modification on github on stable-0.4 and cherry pick into master
19:08 nrzkt      hmmm
19:08 nrzkt      i found a throw which isn't catched on my server
19:09 nrzkt      Server::step => throw ServerError . This throw isn't catched by parents
19:11 nrzkt      i don't know if someone removed the try...catch somewhere but this is triggered by Lua here
19:13 est31      nrzkt, servers dont have catch loops
19:13 est31      they just crash in that case
19:13 nrzkt      yes
19:13 est31      which is bad ofc
19:13 nrzkt      i make an assert
19:13 est31      ummm dunno if zeno will like that
19:14 nrzkt      i do an errorstream and an assert(false).
19:14 nrzkt      it's due to a mod which crash my server
19:14 nrzkt      19:11:09: ERROR[main]: ...st/games/minetest_game/mods/mg_villages/map_of_world.lua:157: bad argument #1 to 'ipairs' (table expected, got nil)
19:14 nrzkt      but the error is not shown because we throw in the nether :p
19:15 nrzkt      est31 i committed your change and uploaded the updated apk to google play store. I'll do it on github when i fix this throw problem
19:17 nrzkt      thanks for the fix
19:18 est31      if you have time, can you review #2460 ?
19:18 ShadowBot  https://github.com/minetest/minetest/issues/2460 -- Optimize minetest.get_(all_)craft_recipe(s) by ShadowNinja
19:18 est31      np
19:19 nrzkt      not atm, i fix the throw problem.
19:19 nrzkt      if(async_err != "") {
19:19 nrzkt      errorstream << "UNRECOVERABLE error occurred. Stopping server. "
19:19 nrzkt      << "Please fix the following error:" << std::endl
19:19 nrzkt      << async_err << std::endl;
19:19 nrzkt      assert(false);
19:19 nrzkt      }
19:19 nrzkt      replacing the throw by this, is this reasonable ?
19:21 nrzkt      this async_err comes from emergethread, which trigger this error from Lua or if network server port is already used
19:23 nrzkt      i don't think there is other design for this. We have the error, we must show it and assert because there is a problem on network or in the Lua code. I commit that to master
19:24 nrzkt      atm we only have a throw segfault without knowing the error
19:28 est31      lol "Please fix the following error:" truly open source
19:29 nrzkt      in fact this will add a Lua backtrace or a port binding error
19:29 nrzkt      then user must fix its addons or network :p
19:31 nrzkt      s/addons/mods/