Time Nick Message 11:23 Krock contains_item is faster than dumping & string find 11:23 Krock wrong channel 13:59 Krock rebased #7569 13:59 ShadowBot https://github.com/minetest/minetest/issues/7569 -- Send custom floats (log2) by SmallJoker 14:01 Krock ^ should be considered for 5.0 to finally have a proper network forat 14:01 Krock *format 14:02 Krock backwards compatibility is rather annoying in this case - due to the resulting additional overhead to decide which type it has to send 14:11 nerzhul Krock for me it's a blocker for 5.0 14:11 nerzhul it's the last chance before many time to fix it 14:11 nerzhul backward compat has no sense on the network part, the min protocol version is already raised 14:20 Krock Right, it would be the right chance to do it 15:35 hecks so I found the offending function for my light bug 15:36 hecks unsurprisingly, the player simply has a half-node Y bias added to the light query position, and other objectrefs don't 15:37 rubenwardy ? 15:37 Krock -> new object property 15:37 hecks now if someone would be kind enough to tell me how do I add a lua property for this, and read it in said function 15:38 Krock https://github.com/minetest/minetest/blob/master/src/object_properties.cpp 15:38 Krock https://github.com/minetest/minetest/blob/master/src/script/common/c_content.cpp#L184 15:38 Krock https://github.com/minetest/minetest/blob/master/src/content_cao.cpp#L724 15:39 hecks the content_cao.cpp bit I've already found 15:39 Krock I think that's all. Only docs would be needed 15:39 hecks I'll see what I can hack here 15:39 Krock should be fairly easy compared to the other bugs 15:39 hecks which is why I picked it 16:15 hecks does adding a new object property mean that the protocol version will need a bump? 16:20 rubenwardy No 16:23 Krock add a try catch around 16:23 Krock or read garbage from the network packet 18:47 hecks are build times always this horrible or would it be better on an actual machine 18:47 Krock make -j4 18:48 Krock ^ could make use of 4 cores at best 18:49 hecks thanks 18:49 hecks from my experience the optimal thread count for a program isn't exactly the number of CPUs, but more if there's any IO overhead 18:50 hecks about 1.5x on average 18:50 p_gimeno I believe 'make -j' should use as many cores as you have 18:51 hecks man says it simply won't limit the number of jobs, so who knows 18:51 Krock ($nproc) or something like that would work too IIRC 18:52 hecks oh I see now, the script had it pegged at 2 jobs 18:53 sfan5 make -j will continue spawning processes until either 1) it's finished or 2) you run out of memory 18:53 Krock I don't see profit there 18:54 Krock * make -j$(nproc) 18:54 hecks so the manpage tells it true then 21:05 hecks Is the name light_offset used for something? 21:06 hecks because my thing broke when I used that exact name for the light anchor thing 21:07 hecks using light_offset made every object fullbright, but light_pos is fine 21:12 hecks okay, never mind, that's not it 21:46 hecks https://a.uguu.se/uk3XUmMKROk3_fixed.png 21:47 hecks had some trouble with BS coordinates but it's all good now 21:48 hecks the new property is called light_anchor and it's in the object's local space 21:48 nerzhul sfan5 did you see that: https://www.zdnet.com/article/windows-apps-made-on-linux-hit-by-security-fail/ ? 21:48 hecks default is 0.5 Y just like the old hardcoded offset for the player 21:55 hecks also, you can force ASLR on windows 21:56 nerzhul not if the program doesn't support it 21:57 hecks the workaround seems simple enough 21:57 nerzhul simple but not user friendly :) 22:00 hecks I mean the code patch in the advisory linked on that page 22:00 hecks just add the line, I guess 22:00 hecks it does make a slightly bigger executable, beyond that I'm not sure how to check 22:00 hecks basically they say to add __declspec(dllexport) before main 22:13 hecks but it doesn't seem to work, procexp says aslr is still off