Time Nick Message 00:01 ShadowNinja Can this be merged? https://github.com/minetest/minetest/pull/656 00:01 ShadowNinja I can't realy develop any more if I can't compile. :-( 00:04 kahrl PilzAdam: if it is the font related segfault (which I'm pretty sure it is), it's unrelated to the F10 segfault 00:04 PilzAdam yea, thought about that too 00:05 PilzAdam why do the delte font thing crash for some people? 00:05 kahrl the font should only be deleted if the code created it 00:05 kahrl that is when freetype is enabled 00:06 kahrl otherwise the font comes from IGUIEnvironment::getFont() and those should not be deleted 00:07 kahrl actually, since it's a subclass of IReferenceCounted, s/deleted/dropped 00:08 ShadowNinja I got minetest to compile! Now lemme see about that backtrace. 00:08 kahrl ShadowNinja: with or without freetype? 00:09 ShadowNinja Without. I will try with. 00:09 kahrl Try both if you have the time 00:11 PilzAdam I can reproduce it with DENABLE_FREETYPE=0 00:12 PilzAdam at /home/adam/Minetest/minetest/src/guiChatConsole.cpp:111 00:13 kahrl yeah when you start the client the second time it does m_font = getFont(...) 00:13 kahrl which returns the exact same pointer as it did the first time 00:13 kahrl (since IGUIEnvironment thinks the pointer is still live) 00:14 kahrl but in fact it has been deleted when the first client quit 00:14 PilzAdam yea, #if USE_FREETYPE arround the delte font in main.cpp and ChatConsole.cpp also fix the crash when shutting down 00:14 ShadowNinja Is anyone working on all of those compile warnings? It is at the point where it is hard to tell if you have added one in the flood of them. 00:15 ShadowNinja If not I might. 00:15 PilzAdam most of them are from mapgen 00:15 PilzAdam kahrl, Ill push that fix now 00:15 kahrl PilzAdam: also replace the delete font with font->drop(); 00:15 kahrl doesn't really matter but that's what you're "supposed" to do 00:16 ShadowNinja PilzAdam: Yes, I know. Is anyone*cough*hmmmm working on them? 00:16 kahrl I sometimes get two warnings per source file from the fwrite calls in debug.h 00:17 kahrl But sometimes I don't 00:17 kahrl My compiler seems confused about whether it should show them or not 00:19 ShadowNinja Why am I still seing the "Server -!- " prepend on massages that shouldn't have it? It worked when I pushed it. Has anyone changed anything around there? Or am I missing something? 00:21 PilzAdam kahrl, done 01:54 ShadowNinja Sapier's new menu will cause some merging dificulties on the IPv6 pull request. Is there any reason not to merge that? 09:53 kaeza can we ban YoyoBuddy already? 09:54 Calinou for what? 09:55 Calinou also this is #minetest-dev 09:55 kaeza offensive PMs 09:55 kaeza oh lol 09:55 kaeza sorry 09:55 Calinou kaeza: screenshot please 18:26 hmmmm alright, just rebased 18:26 hmmmm i hope you're not really going to leave the minetest button as text that says "minetest"... that's going to be the minetest logo, right..? 18:27 PilzAdam it already is... ? 18:27 hmmmm to me it looks like an irrlicht button with the text "minetest" on it 18:27 PilzAdam https://github.com/minetest/minetest_game/commit/7f765f5f6c033ef4123a6fedd726452ad3fb1631 18:27 hmmmm oh, i have to update the game too 18:38 hmmmm hmm pilzadam, now when updating, do i delete common? 18:41 PilzAdam no 18:41 PilzAdam we havent dropped it yet 18:49 hmmmm someone seriously needs to fix the gold texture colors 18:49 hmmmm ugh 18:50 hmmmm also when ice breaks, perhaps it should turn into a water source like in minecraft 18:50 PilzAdam why would ice melt if you hit it with a pickaxe? 18:51 hmmmm why would ice disappear? 18:51 PilzAdam it doesnt disappear, you have it in your inventory 18:52 hmmmm also the snow node is darker than the top tile of snow_on_dirt 18:52 PilzAdam thats an issue with lighting of nodes with special drawtyes 18:52 PilzAdam (nodebox in that case) 18:53 hmmmm that ought to be the next thing to fix i suppose 18:55 hmmmm ahh loving the new sand texture 18:55 hmmmm it actually looks like a beach 18:58 hmmmm deserts actually look like deserts 19:05 VanessaE hmmmm: I provided better textures for the various metal blocks a while back. No one wanted them I guess 19:06 PilzAdam -better 19:06 VanessaE ice should turn into broken ice maybe e.g. a pile of small ice cubes 19:07 VanessaE the lighting issue affects all objects in the game that aren't whole cubes, btw. 19:07 VanessaE signlike, raillike, nodeboxes, etc. 19:15 VanessaE PilzAdam: maybe not better to you, but still they're better. 19:15 VanessaE you aren't the be-all, end-all judge of art :) 19:15 PilzAdam you are neither 19:16 PilzAdam so you are also not qualified to say its better 19:16 VanessaE nope, but I'm a good judge of my own creations :) 19:16 PilzAdam oh, yea, people tend to judge their own things completly neutral 19:17 VanessaE :P 19:52 VanessaE *sigh* more confusion about what has or hasn't got flowers. Why are we waiting on re-merging minetest_game+common+survival+build ? 19:52 PilzAdam dunno 19:52 PilzAdam I could do it now 19:53 VanessaE ok :) 19:56 PilzAdam so, is there any good reason to not do it now? 20:01 VanessaE I don't see any reason not to, as long as everyone's properly informed. 20:01 VanessaE a message should be posted to the forums. 20:02 VanessaE guess I have to tweak my build script again ;) 20:45 hmmmm you mean the realistic texture pack ones? 20:45 hmmmm at the very least, the *colors* used for those are better than what is there now 20:47 hmmmm anyway i guess it would make more sense if ice broke into a water source if we had an equivalent to silk touch in minetest.. 20:47 hmmmm let's say we have an ice sheet in the ocean and somebody digs the ice from the inside out 20:48 hmmmm the water from the rest of the ocean will flood only to a certain point 20:48 hmmmm and then you'll have an interesting hole in the water 21:49 sapier is anyone aware that some irrlicht functions modify current locale setting? 21:53 hmmmm i am now.... which ones? 21:53 hmmmm and why 21:54 sapier haven't exactly figured out but seems to be related to file open/save dialogs 21:55 sapier this results in atof returning different values prior opening a dialog than after opening 21:55 hmmmm sorta wonder if this is the cause of any bugs we've seen 21:55 sapier as atof uses current locale setting to parse a float value ... I don't think we realy want formspecs to be dependent on current locale setting ... do we? 21:56 hmmmm is there a locale independent variant of atof? 21:56 hmmmm like some sort of atof_l()? 21:56 sapier i don't think so but we can set a defined locale prior using atof reverting it after it ... quite a hack 21:56 sapier but atof already is hacky 21:57 sapier it doesn't do any error handling 21:57 hmmmm i worry about the speed 21:57 hmmmm atof is very hot the way it is 21:57 hmmmm sure it's not some sort of killer, but it's a performance degredation 21:57 sapier yes but speed is nothing if result is not reliable 21:58 hmmmm so let's change the locale and back every single time we get a setting 21:58 sapier of course I'd prefere to have a fast and good solution too 21:58 sapier it's not only setting, it's on each redraw of a formspec 21:58 hmmmm right, that too 21:59 sapier a own implementation would be an option ... still I don't know how much performace difference would be 21:59 hmmmm hah 21:59 hmmmm have you seen the bsd implementation of atof()? 22:00 hmmmm 1030 lines 22:00 sapier what? 22:00 sapier assembler code? 22:00 hmmmm C code 22:01 sapier what are they doing? 22:01 hmmmm i have not the slightest clue 22:01 hmmmm i'd pull it up, but i can't be arsed 22:02 hmmmm just saying, atof() is probably sucky the way it is, and if we do write our own, there's a good chance it'd be better 22:02 sapier maybe another solution would be setting locale to a defined value on minetest startup 22:02 hmmmm what about istream 22:02 hmmmm in minetest a lot of strings are parsed with istream, you know 22:02 hmmmm and i bet they call atof(), not their own thing 22:02 hmmmm so all that is affected as well 22:02 sapier there's a comment in mystof that stringstream would crash on mingw ... maybe due to missing locale setting? ;-) 22:03 hmmmm ohh that retardation 22:03 sapier missing locale setting is just a guess :-) 22:03 hmmmm yeah celeron never figured out what the problem with that was, mingw is just bugged 22:04 hmmmm erm so wait, what's wrong with simply changing the locale back when an irrlicht dialog is finished? 22:04 sapier suggestion for using istringstream is calling std::locale("C") ;-) ... not much of a difference to setting directly :-) 22:04 hmmmm also, before you do anything radical, it might be best to check if irrlicht has some kind of setting to disable this auto-locale-changing 22:05 sapier tried this it doesn't seem to work 22:05 hmmmm it's probably good to know why it doesn't work... 22:05 sapier looking a the locale to be set I guess setlocale refuses to set locale back to undefined value :-) 22:07 sapier maybe setting locale to defined value on minetest startup even stops irrlicht from setting a own locale value ... I'll try 22:07 hmmmm allrighty 22:07 hmmmm and i'll finally fix the mapgen memory leaks/warnings 22:08 hmmmm everybody bitches about those constantly... i don't even get any warnings 22:08 sapier great :-) 22:08 sapier wait till ppl want mapgen to create mountains and flatlands .. of course same time at same location ;-) 22:08 init sapier: random question offtopic to this, can entities auto-despawn? 22:09 sapier no 22:09 sapier you have to handle this within lua code 22:09 init hmm, k, thanks 22:10 sapier e.g. count some internal value up on_step and make the entity remove itself after some value is reached 22:10 init i really want to setup a reduced-spawn-ratio simple_mobs_version, but as they actually don't automagicaly despawn, i'm going to leave it out for now 22:11 hmmmm uh oh, i am reading pilzadam's fix for this and he tries to free globals 22:11 hmmmm welp, good thing i told him to wait 22:12 PilzAdam I actually had some ifs to prevent it 22:12 hmmmm i see that, i'm talking about indev though. 22:12 PilzAdam indev doesnt have predefined global constants 22:13 hmmmm nparams_indev_def... 22:13 hmmmm which seems like it's never even assigned anything 22:14 PilzAdam https://github.com/PilzAdam/minetest/blob/master/src/mapgen_indev.cpp#L27 22:14 hmmmm omg serious QC needed in the indev 22:14 hmmmm yes, maybe the dangerous code path isn't ever executed in practice, but still... 22:15 sapier init ... use mobf ... this issue is fixed there ... simple mobs is simple so once you start adding all this workarounds to it you're better up with mobf 22:15 init would it "make them" despawn when the chunks get unloaded? 22:15 PilzAdam noo! dont use mobf, its bloated ;-) 22:16 init mobs war! 22:16 PilzAdam init, despawning on deactiavation of the block is not supported 22:16 init PilzAdam is the stone monster and sapier the vombie(?) :P 22:16 init hm 22:16 init that is the only thing i wanted :P 22:16 PilzAdam there is a pull request, IIRC 22:16 sapier PilzAdam he wants to bloat simplemobs to so he can use mobf ;-P 22:17 hmmmm PilzAdam, have you seen any virtual deconstructor warnings related to the Ore code? 22:17 PilzAdam not that I can remember 22:17 hmmmm whatt 22:17 hmmmm why _isn't_ there 22:18 hmmmm because i don't have any base destructor 22:18 hmmmm so there's a memory leak 22:18 hmmmm how come OreSheet never showed up as a leak in valgrind then? 22:18 PilzAdam I guess because it isnt used? 22:19 hmmmm i guess 22:20 PilzAdam oh, yes I get warning about Ore contstructors 22:20 hmmmm and not destructors? 22:20 PilzAdam /home/adam/Minetest/minetest/src/scriptapi.cpp:691:10: warning: deleting object of abstract class type 'Ore' which has non-virtual destructor will cause undefined behaviour [-Wdelete-non-virtual-dtor] 22:20 hmmmm figured as much 22:20 hmmmm it would only make sense, i mean 22:21 PilzAdam there simply such a flood of the Noise warnings that I havent seen it until now 22:21 hmmmm if my compiler had that warning this wouldn'tve ever been an issue even 22:22 hmmmm Noise warnings? you mean NoiseIndev, no? 22:22 hmmmm there is literally no reason why Noise, as a base class, should have any warnings at all 22:23 PilzAdam /home/adam/Minetest/minetest/src/mapgen_v7.cpp:109:9: warning: deleting object of polymorphic class type 'Noise' which has non-virtual destructor might cause undefined behaviour [-Wdelete-non-virtual-dtor 22:23 PilzAdam same for indev and v6 22:24 hmmmm oh 22:24 hmmmm so having a class derived from that turns the base class into a polymorphic class type 22:24 hmmmm in this case, though, i know that the warning is wrong and it wouldn't cause any such UB 22:25 hmmmm but i'll add something in to shut it up regardless 22:25 sapier ok setting locale at startup doesn't change anything 22:27 sapier once storing content of setlocale returned char* restoring does work too ... still if someone happens to set his system numeric locale parsing of formspecs will fail in minetest 22:27 sapier as well as any other string to float operation 22:29 hmmmm so wait, why are formspecs parsed every time they're drawn? 22:29 hmmmm that's hacky and stupid in the first place 22:30 sapier not exactly everytime they are modified 22:30 sapier formspec update/window size change 22:30 sapier so it's not that often 22:33 sapier still the main problem atof beeing locale dependent remains 22:35 hmmmm wait, what did you try setting exactly 22:35 hmmmm setlocale(LC_ALL, "C")? 22:35 sapier yes 22:35 hmmmm er 22:35 sapier irrlicht overwrites it so doesn't make any difference what you set there 22:36 hmmmm you think we should open an issue ticket with irrlicht? 22:37 sapier maybe ... im my case irrlicht repaced a whole bunch of locale settings by main one only 22:38 sapier thus my setting for numeric was lost 22:42 hmmmm still don't see what's wrong with setting the locale back after opening an irrlicht dialog though 22:43 sapier I do it this way ... but I can't tell if this is the only case where it happens 22:43 sapier and of course once a user has some strange system locale minetest will fail too 22:45 hmmmm how do i set what CC to use with cmake? 22:46 sapier I don't have any idea 22:46 PilzAdam CMAKE_CXX_COMPILER 22:47 PilzAdam CMAKE_C_COMPILER 22:47 PilzAdam (see CMakeCache.txt) 22:47 hmmmm great 22:49 sapier hmmmm resetting locale after dialog is problematic either ... the dialog may be open for quite some time so any atof call while a dialog is shown may fail 22:49 hmmmm are the dialogs blocking? 22:50 sapier at most their own thread but there are others 22:50 sapier and in my case not even the own thread is blocked 22:51 sapier currently noone uses irrlicht dialogs so this is not exactly a current problem ... but it may have already happened for users with strange locale settings 22:52 hmmmm let's just not use irrlicht dialogs 22:52 hmmmm there's no good way to fix this, so let's avoid the problem altogether 22:53 sapier once someone volonteers to build a file select dialog on it's own I'll be perfectly fine with this suggestion ;-) 22:54 sapier for what I intend to do fixing after closing dialog will be fine ... but I guess once the dialog is added someone will use it in situations where messing up locale will result in crashes 22:55 hmmmm why am i still not getting the dtor warnings... 22:55 hmmmm is this specific to gcc47? 22:55 hmmmm i am trying with gcc46 and not seeing it 22:55 sapier non virtual destructor? no its in 4.2 too 22:56 * PilzAdam uses 4.7.3 22:56 hmmmm i don't get these warnings at all 22:56 sapier <4.7.2 and 4.6.x 22:56 sapier maybe different on different os? 22:56 hmmmm hold on 22:57 sapier or different default warning levels? 22:57 ShadowNinja 4.8.0, I win. :-) 22:57 hmmmm i didn't change the warning level at all 22:57 hmmmm it must be the OS difference then 22:57 VanessaE I'm seeing those dtor warnings here. (4.7.3) 22:57 VanessaE quite a few of them in fact. 22:57 PilzAdam sapier, doesnt minetest compile with -Wall ? 22:57 sapier lol :-) your bugs won't be accepted shadowninja you most likely have to many new ones ;-) 22:58 sapier pilzadam I don't know if you can specify checks for gcc on compiletime too? 22:59 sapier -if +maybe 22:59 sapier arg 1 o clock am ... gth ;-) 23:00 hmmmm well in order to fix something i need to be able to see it first 23:00 PilzAdam maybe the warnings just hide on hmmmm's computer because they dont want to be fixed :-) 23:01 VanessaE export CXXFLAGS=-O3; cmake . -DCMAKE_INSTALL_PREFIX:PATH=/usr -DRUN_IN_PLACE=0 -DENABLE_FREETYPE=0; make -j7 23:01 VanessaE if any of that changes things. 23:01 hmmmm well for one i can't use export 23:02 VanessaE well, whatever your shell uses in its place. 23:02 hmmmm none of that will help with this though