Time  Nick        Message
02:48 luizrpgluiz hello, I enjoyed the new daily build of minetest here in kubuntu 16.04 beta 2, the game's performance was much better now with disabled shaders
04:42 maxxrodz    hi
04:50 maxxrodz    people are getting ignored
04:50 maxxrodz    is this minetest
04:50 maxxrodz    ok
04:50 maxxrodz    is this minetest
04:50 maxxrodz    hi
04:53 sofar       this channel is for development of the core. other topics, please talk in #minetest, thanks.
19:35 srifqi      ~tell paramat I've updated 2561, please see it.
19:35 ShadowBot   srifqi: O.K.
19:59 red-001     #3919
19:59 ShadowBot   https://github.com/minetest/minetest/issues/3919 -- Add base64 encoding and decoding to the lua api. by red-001
20:03 ShadowNinja I found some issues with the clang static analyzer.  Some of them look like false positives, but it found a few good mem leaks.
20:23 Calinou     meme leaks
20:23 VanessaE    USE UP _ALL_ THE RAM!
20:35 Fixer       VanessaE, develope all the things
22:26 ShadowNinja Clang likes the (new GUIMenu())->drop() pattern just about as much as I do.
22:28 ShadowNinja Apparently set_lists didn't work at all, and safeLoadFile had some leaks.
22:29 ShadowNinja I also found some NULL dereferences and an issue with the SRP API.
22:29 ShadowNinja It handles the case where malloc returns NULL, but then ignores it and operates on uninitialized data.
22:30 ShadowNinja IMO a memory allocation failure should just throw std::bad_alloc and crash, since there isn't much you can do about it.
22:31 ShadowNinja But for not I just made it propagate the error to the caller as a generic SRP error.
22:44 ShadowNinja generateBiomes seems to be copies&pasted across all mapgens but v6.
22:57 ShadowNinja Likewise with checkobject and Lua class-like userdata.
23:28 est31       the ->drop() is utter shit
23:28 est31       because it isn't called when unwinding the stack
23:29 est31       so e.g. if you allocate, and unwind before the drop, it gives big trouble
23:30 est31       ShadowNinja, about SRP, the bad alloc throwing should happen at the border, and on the minetest side of the story.
23:30 PilzAdam    est31, https://github.com/minetest/minetest/commit/25da0594eb09bb0e72816aaf8280573a668af00c#commitcomment-16856039 thanks
23:30 est31       PilzAdam, np
23:31 est31       ShadowNinja, unless there is an actual bug in SRP itself.
23:31 ShadowNinja est31: In that case the object grabs itself and then frees itself when done though, which I think is a pretty ugly design.
23:32 ShadowNinja est31: And as for SRP, minetest will only get "SRP authentication failed", nothing more specific.
23:32 est31       can you show the code
23:33 ShadowNinja est31: hash_num in calculate_M drops the error.
23:34 est31       hmmm
23:34 est31       err
23:34 est31       I meant hrmmm
23:35 est31       hmmmm, sorry didnt want to ping you
23:35 ShadowNinja Compiler warning/bug fixing patch so far: http://ix.io/uZy/diff
23:35 ShadowNinja est31: You didn't -- until you added that extra 'm' :-P
23:35 est31       lol
23:35 est31       well shrug
23:36 est31       ShadowNinja, about the patch, I don't like it. SRP is meant as a C library, the c++ code is only included for the hash alg i think
23:37 est31       throwing bad_alloc is a bad habit of c++ libraries, C libraries just return NULL pointers or otherwise indicate an error
23:37 ShadowNinja est31: What would you prefer?
23:38 ShadowNinja est31: calculate_M is called by srp_user_process_challenge, which doesn't even really have a way to report errors.
23:38 est31       ShadowNinja, make it return an int and check where the function is called
23:39 ShadowNinja s/int/bool/  I did that.  It prevents the read of uninitialized data, but it otherwise just ignores the error.
23:40 est31       ShadowNinja, C has no bool type
23:40 est31       at least last time I checked
23:40 ShadowNinja est31: stdbool.h / _Bool
23:40 ShadowNinja c99
23:41 est31       ah well
23:42 ShadowNinja est31: Line 66/67 of the patch, I'm not sure which time_of_day should be used.  I used the current value, but that may be a bug.  Do you know which is correct?
23:45 est31       I have no idea why you are even changing it
23:45 est31       reviewing this is pointless without having the output from the clang program
23:49 est31       its too large to be reviewed via irc, it best gets in line as a normal PR
23:49 est31       sorry
23:49 ShadowNinja est31: time_of_day is initialized to the struct's value, then immediately reset to the client's value.
23:49 est31       (also partly I want to do other things, non minetest things now)
23:50 ShadowNinja O.K.