IRC logs for #openttd on OFTC at 2020-02-01
⏴ go to previous day
00:03:43 <Wolf01> First lego purchase of the year \o/
00:04:18 <frosch123> Heiki: they are all drunk
01:29:57 *** WormnestAndroid has quit IRC
01:31:08 *** WormnestAndroid has joined #openttd
01:33:16 <DorpsGek_III_> [OpenTTD/OpenTTD] glx22 opened pull request #7967: Fix #7966: SQInteger is 64-bit, print it as 64-bit https://git.io/Jv3gN
01:37:17 <DorpsGek_III_> [OpenTTD/OpenTTD] James103 commented on issue #7966: Error in script error message for type mismatch https://git.io/Jv3fc
02:16:01 <DorpsGek_III_> [OpenTTD/OpenTTD] glx22 commented on issue #7966: Error in script error message for type mismatch https://git.io/Jv3fc
03:53:22 *** snail_UES_ has joined #openttd
05:29:47 *** supermop_work has joined #openttd
05:54:47 *** WormnestAndroid has quit IRC
05:55:31 *** WormnestAndroid has joined #openttd
05:55:41 *** Laedek_ has joined #openttd
06:05:56 *** WormnestAndroid has quit IRC
06:06:09 *** WormnestAndroid has joined #openttd
06:16:18 *** dihedral has joined #openttd
07:58:31 *** tokai|noir has joined #openttd
07:58:31 *** ChanServ sets mode: +v tokai|noir
08:48:31 *** Progman has joined #openttd
08:54:49 *** Extrems has joined #openttd
09:36:07 *** ChanServ sets mode: +v tokai
09:55:36 *** sla_ro|master has joined #openttd
10:04:45 *** Smedles has joined #openttd
10:21:27 *** Smedles_ has joined #openttd
10:38:11 *** andythenorth has joined #openttd
11:14:48 <DorpsGek_III_> [OpenTTD/OpenTTD] LordAro approved pull request #7967: Fix #7966: SQInteger is 64-bit, print it as 64-bit https://git.io/Jv3Xt
11:14:55 <DorpsGek_III_> [OpenTTD/OpenTTD] LordAro merged pull request #7967: Fix #7966: SQInteger is 64-bit, print it as 64-bit https://git.io/Jv3gN
11:59:43 <nielsm> so, how does it feel to have been brexited?
12:01:44 <andythenorth> there were no fireworks last night
12:01:56 <andythenorth> on the other hand...where I live had one of the highest remain votes so eh
12:02:24 <andythenorth> I think it's potato/potato
12:04:45 <andythenorth> shall I do Euro Horse next?
12:10:59 <LordAro> andythenorth: Iron Horse but blue with some yellow bits?
12:11:07 <andythenorth> and no UK trains at all
12:11:15 <andythenorth> also no Norwegian trains
12:13:13 <Pikka> what happened to Amerihorse?
12:14:25 <andythenorth> I want to do them all!
12:14:38 <andythenorth> and FIRS 4, and Unsinkable Sam, and Road Hog 2
12:14:42 <andythenorth> and fix the newgrf docs
12:16:11 <Pikka> I think I'm going to go back to rendering for av9
12:16:50 <Pikka> even with the av8 sprites as a guide, aircraft are just too complicated shapes to hand-draw all the directions for. Never mind tilted sprites for taildraggers or takeoffs
12:17:05 <andythenorth> Sam is stuck waiting for me to draw some hulls
12:17:13 <andythenorth> right pain in the arse
12:17:46 <andythenorth> NARS 9000 Horse, I never figured out the gameplay progression
12:17:57 <andythenorth> as US trains have got progressively slower since the 1930s
12:18:06 <andythenorth> which isn't what games do :P
12:20:34 <Pikka> take the inspiration from what fits the gameplay, then. that's why I left out the early pacifics from UKRS3
12:21:32 <Pikka> the way I've mapped out NARS 2001 is a progression of more powerful, but slow, freight locos from 1900 - conso, mikado, mallet or something else big
12:23:16 <Pikka> and fast-ish, but not too big passenger locos - American, Pacific, etc. Then it all gets replaced by much less powerful, but slightly faster freight diesels which get MU'ed, and 100mph passenger locos. I guess it would play similar to the UK set, although "powerful" is a lot *more* powerful for the US
12:24:02 <Pikka> and anything "realistic" that doesn't fit, like giant express steam locos or centennials etc gets ignored :P
12:29:12 <andythenorth> I am pretty determined to include some stupid things like the UP turbines
12:29:19 <andythenorth> which also only last 15 years :P
13:35:52 *** andythenorth has joined #openttd
14:03:26 *** WormnestAndroid has quit IRC
14:03:40 *** WormnestAndroid has joined #openttd
14:18:37 *** WormnestAndroid has quit IRC
14:18:55 *** WormnestAndroid has joined #openttd
14:22:55 *** snail_UES_ has joined #openttd
15:17:04 *** frosch123 has joined #openttd
15:34:13 <_dp_> meanwhile I'm deep into a wtf land with my sprite sorters as I've no longer have any idea why some optimizations work and some don't xD
15:35:07 <_dp_> on a positive side, they keep getting faster xD
15:36:10 <glx> it probably depends on the initial state
15:39:15 <_dp_> glx, initial state of what?
15:39:43 <glx> dunno, almost sorted or highly unsorted
15:39:46 <_dp_> it probably depends on some properties of input data that I don't quite understand
15:40:03 <_dp_> as they are already way beyond theoretical limits for worst case performance
15:41:00 <_dp_> it's almost sorted by definition, otherwise original sorter would never finish xD
15:41:27 <glx> maybe the sprite sizes too
15:42:06 <glx> if you don't have a big sprite hidding many others it's probably the worse case
15:43:27 <_dp_> sprites are all relatively small in openttd
15:44:09 <_dp_> it's some really weird properties. for example if I take sorter in that PR and use x+y as order instead of x+y+z it becomes twice as fast xD
15:45:02 <_dp_> and z is rarely above 100, how on earth does that even matter %
16:31:01 <andythenorth> I don't recall :)
16:31:12 <andythenorth> I don't really trust my local tests tbh
16:31:24 <nielsm> you are reproducing something
16:31:27 <andythenorth> the performance self-built is so much worse than the official binary
16:31:38 <andythenorth> and there seems to be no way to address that
16:31:59 <nielsm> it's still not so bad that you should be getting 17 fps
16:32:22 <nielsm> if the framerate window says the total game loop time is below 20 ms then you should be running at a constant 33.3 fps
16:33:16 <nielsm> and if the framerate is significantly worse than that, but the game loop time is not in the red then that in itself is the bug
16:34:01 <nielsm> you have that thing where if you show the vehicles list the framerate drops massively, right?
16:34:25 <nielsm> have you tested whether making the game even more complex causes the framerate to drop even more?
16:34:27 <DorpsGek_III_> [OpenTTD/OpenTTD] glx22 updated pull request #7270: Introduce CMake (and removing all other project-related code) https://git.io/fhbqc
16:34:37 <nielsm> e.g. taking the game loop time from 5 ms to 20 ms
16:34:42 <andythenorth> maybe there's a save we could use as a baseline?
16:35:33 *** snail_UES_ has joined #openttd
16:35:34 <glx> #7904 may be the cause for windows with lists
16:36:31 <nielsm> that's likely one bad thing yes
16:37:21 <andythenorth> I am curious why the binaries built on the CF are so much faster
16:37:49 <nielsm> okay that's probably about the same speed as mine, i5-4460 3.2 ghz (mine is 4th gen, yours is 8th gen)
16:40:50 <andythenorth> hmm software update BIAB
16:43:17 <nielsm> ProZoneGame 21 runs just on the edge of 33 fps for me, that might be a good test?
16:43:54 <nielsm> maybe it's also worth trying to load a game in dedicated server mode, so you can measure without graphics
16:44:05 <nielsm> and see if the graphics driver used makes a difference
16:44:16 <frosch123> you can also just pause the game
16:44:35 <frosch123> i used some coop game and compared paused vs unpaused
16:44:59 <nielsm> frosch123 nah the issue is also with simulation speed affecting things in weird ways
16:45:25 <nielsm> it seems when the game loop time crosses some threshold, the total framerate drops much more than it should given just the game loop time
16:46:17 <glx> well if gameloop takes too much it triggers hundred ticks more often I think
16:46:44 <nielsm> it's not on that level
16:47:21 <nielsm> we're talking game loop + graphics painting going from 2 ms to 5 ms causing the framerate to drop from 34 to 17
16:47:43 <nielsm> 17 should only be happening if the game loop time + drawing time was around 60 ms
16:48:35 <glx> oh like if it a little above 30ms it's like 60ms ?
17:03:56 <glx> OSX doesn't seem to use threaded drawing (but my search in the code may be failing :) )
17:05:59 *** andythenorth has joined #openttd
17:07:22 <andythenorth> so how do I replicate a production build (instead of debug)?
17:07:33 <andythenorth> I tried "./configure --enable-debug=0 --disable-assert"
17:07:52 <glx> you can check how it's done in azure files
17:11:20 <andythenorth> hmm prozone game has unavailable grfs
17:12:07 <andythenorth> nielsm: we need an agreed method for reporting fps
17:12:13 <glx> andythenorth: in the logs for nightly I see "./configure PKG_CONFIG_PATH=/usr/local/lib/pkgconfig --enable-static"
17:12:51 <LordAro> only releases have asserts disables
17:14:40 <nielsm> there's a way to make things slow without needing specific saves
17:14:47 <nielsm> at least it works for me :)
17:15:07 <andythenorth> I'll try it shortly
17:15:20 <andythenorth> so much affects the reported fps
17:15:25 <andythenorth> including the fps window itself
17:15:39 <andythenorth> animation on/off
17:16:19 <glx> I think tests should be done with animation off ;)
17:16:57 <nielsm> I'd say test with slowdown on a blank map in editor
17:17:20 <andythenorth> I'm only now reporting fps with the windoshade
17:17:27 <andythenorth> the drawing of text is so parasitic
17:19:24 <andythenorth> I've turned all news messages off, they're a horrible performance suck afaict
17:19:27 <andythenorth> including the summaries
17:20:29 <nielsm> it should be possible to set slowdown such that the game loop time constantly skirts 30 ms, and still get 32-33 fps
17:22:07 <andythenorth> ok, with the configure command glx posted above, my self-built binary is now marginally faster at ffwd than official build
17:23:07 <glx> hmm --enable-debug is always a debug build anyway
17:23:36 <glx> with more or less optimisations but still a debug build I think
17:29:47 <frosch123> LordAro: the "--disable-assert" was added for releases only by the old compile farm
17:30:04 <frosch123> we added it in the 1.9 branch, but auto-detetion was postponed for cmake
17:32:14 <frosch123> actually... it wasn't the compile farm that added it, but it was a diff between svn-branch and svn-tag
17:40:32 <andythenorth> oh hover rows in saveload :)
17:41:28 <andythenorth> nielsm: so your patch in the gist...I've applied it
17:41:59 <andythenorth> doesn't seem to reduce fps though :)
17:42:12 <nielsm> the slowdown command has no effect?
17:42:28 <nielsm> then it's probably clang being too smart about optimisation :D
17:42:49 <LordAro> assuming andythenorth has actually run the slowdown command...
17:47:10 <andythenorth> command is just 'slowdown'?
17:47:52 <nielsm> for me "slowdown 21" gives me a game loop time of 30 ms
17:49:02 <andythenorth> slowdown 21 is giving avg game loop time 50ms
17:49:16 <nielsm> try smaller values until you hit about 30 ms
17:49:33 <andythenorth> hmm the average declines
17:49:49 <andythenorth> ok I just leave it a minute or so to settle
17:50:20 <nielsm> the important part is, what overall fps do you get when the game loop time is just about 30 ms?
17:51:56 <andythenorth> most common value is around 19.6fps
17:52:10 <andythenorth> it's typically 19-20fps
17:53:20 <nielsm> try lowering the slowdown by 1 step at a time, and see if there's any points where the fps makes sudden jumps to a different range
17:53:25 <nielsm> or if the fps changes smoothly
17:58:20 <frosch123> i also don't get why the screenshot tool adds a border of transparency
17:59:10 <nielsm> I don't get any spiking with the slowdown
17:59:33 <nielsm> ♬ John Broomhall - Transport Tycoon Deluxe (MT-32) - Jammit (3:34 @ 240 kbps) ♬
18:04:03 <andythenorth> ok at slowdown 15, I initially get the spikes
18:04:13 <andythenorth> then the game runs at 22-23fps
18:04:41 <andythenorth> at slowdown 16, it's 19-20fps (this is the 30ms game loop)
18:04:56 <andythenorth> slowdown 15, the game loop is about 17ms
18:06:32 <andythenorth> yeah, ignore that
18:06:40 <andythenorth> I opened the game loop chart
18:06:45 <andythenorth> that completely changes performance
18:07:02 <andythenorth> oof this is confusing :)
18:07:22 <andythenorth> opening the game loop chart nearly halves game loop time
18:08:13 <nielsm> something strange seems to be going on with windows that update on a timer then?
18:08:33 <andythenorth> much as the fps reporting is necessary and cool
18:08:50 <andythenorth> it's got a real measurement-changes-the-measurement problem
18:17:02 <andythenorth> there's some weird non-deterministic choking in the middle of that video
18:17:05 <andythenorth> that doesn't usually happen
18:20:22 <nielsm> btw I'd suggest testing with the scenario editor on an empty map
18:20:32 <andythenorth> spukhafte fernwirkung
18:20:36 <nielsm> (with no newgrf, ai, gs loaded)
18:43:12 <nielsm> ...so actually I do have a 2010 macbook pro around here but I have no idea if it works, at least I know the SSD is dead
18:43:34 <nielsm> but I wonder if it'd be possible to install current macos on it?
18:46:22 <frosch123> nielsm: when i replace the getstring loop with a usleep, i get no spikes
18:47:27 <nielsm> but I wanted to avoid sleep since that calls into the os and typically deschedules the thread
18:47:47 <nielsm> and when part of my suspicion lies with sleep calls and/or the thread scheduler that's not something I want
18:48:45 <frosch123> i am not sure usleep is a call to the os
18:53:50 <nielsm> hm well, the charger works for this, getting the orange charging led
18:58:22 <andythenorth> if the SSD is dead, USB external? o_O
18:58:33 <andythenorth> ouch, probably just USB 1.1 or something :P
18:59:12 <nielsm> it's so old it uses 2.5" harddrives
18:59:20 <nielsm> and I have a spinny disk installed in it now
18:59:59 <nielsm> the built-in recovery thing does boot, but when I try to have it reinstall osx it fails during the "verifying elibility with apple" initial step
19:00:05 <nielsm> and it does have internet connection too
19:00:15 <nielsm> I guess apple doesn't serve 10.8 any longer
19:02:41 <andythenorth> allegedly "El Capitan or other installs will show corrupted if the time and date on the machine is after the 10 January 2016, just change it and it will install."
19:04:34 <DorpsGek_III_> [OpenTTD/DorpsGek] pyup-bot opened pull request #20: Scheduled monthly dependency update for February https://git.io/Jv3bw
19:11:32 <DorpsGek_III_> [OpenTTD/OpenTTD] JGRennison commented on issue #6566: Very long loading of the maximum "zoom out" level in 4K resolution https://git.io/fxABA
19:21:54 <DorpsGek_III_> [OpenTTD/OpenTTD] JGRennison opened pull request #7968: Fix #6566: Fix signed integer overflow in viewport draw area chunking https://git.io/Jv3Nt
19:38:04 *** Fahrradkette has joined #openttd
19:41:21 <andythenorth> involves deleting and rebuilding the station until it's in the right place
19:41:35 <nielsm> sometimes you can remove a station part and add it back
19:41:38 <andythenorth> alternatively, isn't it fixed in 1.10 betas?
19:41:46 <nielsm> or add an extra station part and remove that again
19:41:55 <nielsm> yes it exists only in 1.9.3 and no other versions
19:42:16 <Fahrradkette> so it centers when expanding/shrinking the station?
19:42:21 <nielsm> the bug was introduced because I made a mistake when merging one patch into the 1.9 branch
19:42:47 <Fahrradkette> or could I bin-patch a savefile?
19:42:48 <nielsm> nah, it moves to the top corner if it ends up outside the station bounding rectangle
19:43:52 <Fahrradkette> wow, that worked, thanks nielsm :)
19:45:40 <Fahrradkette> btw, when will 1.10 be released?
19:46:33 <nielsm> uh, maybe the usual april 1st, not decided yet
19:46:45 <nielsm> we need to get around to make a release candidate soon
19:47:17 <Fahrradkette> good to hear, looking forward to it :)
19:48:02 <Fahrradkette> 1st of april is imho a good release date...if stuff breaks one can say "april's fool!" :P
19:53:00 <milek7> slowdown behaves rather weird
20:02:14 <frosch123> milek7: enable fast forward
20:03:12 <frosch123> then you can see more gameticks, so you get a better average
20:04:31 <frosch123> noone knows, i got a lot less jumps when replacing the loop with usleep
20:04:47 <frosch123> computers are complex machines, you never know which butterfly does what
20:04:50 <milek7> btw. graph is somewhat broken with high fps
20:05:14 <frosch123> graph's autoscaling is bonkers
20:05:40 <frosch123> otherwise it just jumps around all the time
20:06:04 <milek7> uh, imgur lately started to compress images too much
20:06:11 <nielsm> okay maybe tls was breaking a bit here, the RTC had lost power and reset to dec 31st 2000
20:21:02 *** tokai|noir has joined #openttd
20:21:02 *** ChanServ sets mode: +v tokai|noir
20:40:26 *** snail_UES_ has joined #openttd
21:15:45 <DorpsGek_III_> [OpenTTD/OpenTTD] James103 opened issue #7969: Console command "alias" can stack overflow and crash (bypassing crash handler) https://git.io/Jv3xF
21:15:59 <DorpsGek_III_> [OpenTTD/OpenTTD] Hezkore commented on issue #7889: [Request] Company rating via Server Admin Port https://git.io/JeAhZ
21:36:10 *** agentw4b has joined #openttd
21:42:27 <DorpsGek_III_> [OpenTTD/OpenTTD] glx22 commented on issue #7969: Console command "alias" can stack overflow and crash (bypassing crash handler) https://git.io/Jv3xF
21:43:40 <DorpsGek_III_> [OpenTTD/OpenTTD] LordAro commented on issue #7969: Console command "alias" can stack overflow and crash (bypassing crash handler) https://git.io/Jv3xF
21:44:17 <DorpsGek_III_> [OpenTTD/OpenTTD] glx22 commented on issue #7969: Console command "alias" can stack overflow and crash (bypassing crash handler) https://git.io/Jv3xF
21:44:27 <DorpsGek_III_> [OpenTTD/OpenTTD] nielsmh commented on issue #7969: Console command "alias" can stack overflow and crash (bypassing crash handler) https://git.io/Jv3xF
21:56:09 *** Lejving has joined #openttd
22:10:24 <LordAro> and wouldn't reset for successive alias calls
22:45:34 <glx> updated with something cleaner I think
23:00:41 <nielsm> I'd change the error message to something slightly less computer sciency, "Too many alias expansions, aborting." maybe
23:01:14 <nielsm> or "Too many alias expansions, recursion limit reached. Aborting."
23:18:38 *** andythenorth is now known as Guest15134
23:18:46 *** andythenorth has joined #openttd
23:19:04 *** andythenorth has left #openttd
23:24:11 *** gelignite has joined #openttd
23:37:11 *** supermop_Home has joined #openttd
23:49:22 <TrueBrain> "I tried to follow aliases for more than ten times. Aborting" or something. Naming the limit and avoiding 'recursion' would help for the average Joe I think
23:50:22 <TrueBrain> "Alias calling an alias calling more aliases .. and this for more than ten times. Aborting"
23:58:07 <LordAro> it is a console command
23:58:13 <LordAro> doesn't need to be totally average Joe
23:59:19 <TrueBrain> Small effort, huge gain ;)
continue to next day ⏵
