02:12 <@heinrich5991> deen: TSC? 02:15 <@deen> time stamp counter 02:16 <@deen> fastest clock source by far on x86-64, just a cpu instruction, 40 ns, but broken on my system after suspend 02:17 <@deen> manwhile hpet takes 4000 ns... 02:28 <+Neruell> objs/base/system.o: In function `time_get': 02:28 <+Neruell> system.c:(.text+0x546): undefined reference to `clock_gettime' 02:28 <+Neruell> collect2: ld returned 1 exit status 02:28 <+Neruell> bam: 'DDNet-Server' error 256 02:30 <@deen> Oh, we need to link with -lrt for old glibc versions 02:31 <@deen> thanks Neruell 02:31 <+Neruell> np 02:32 <@deen> Please try again 02:33 < ddnet-commits> [ddnet] def- pushed 1 new commit to master: https://git.io/vobfz 02:33 < ddnet-commits> ddnet/master 3c7a9bc def: Link with -lrt for clock_gettime for glibc < 2.17 02:36 <+Neruell> gj it works 02:36 <+Neruell> :P 02:40 <@deen> the more difficult test is whether this breaks anything 02:40 <@deen> prediction, physics, animations 02:44 <+Neruell> why do u use CLOCK_MONOTONIC? 02:45 <@deen> instead of? 02:51 <@deen> the problem with gettimeofday was that it jumps around when you change the system time 02:51 <@deen> CLOCK_MONOTONIC doesn't, so we can now use ntp on servers without having to worry about that 02:52 <+Neruell> i prefer CLOCK_MONOTONIC_RAW 02:53 <@deen> it's linux specific 02:53 <@deen> oh wait 02:53 <@deen> from what I read in clock_gettime CLOCK_MONOTONIC is affected by NTP... 07:19 <+minus> It is? I thought the point of it was that it isn't 11:36 <@heinrich5991> minus: it is, but it is good that it is 11:36 <@heinrich5991> it basically makes 1 second one second even if the hardware doesn't 11:37 <@heinrich5991> do that correctly 11:37 <@heinrich5991> it's not affected by clock jumps though 11:57 <+ScottKevill> It can be affected by NTP while still being monotonic. 12:07 <@heinrich5991> yes