Log for #openttd on 9th October 2021:
Times are UTC Toggle Colours
00:08:47  *** HerzogDeXtEr has quit IRC
01:36:46  *** Gustavo6046 has quit IRC
01:37:20  *** Gustavo6046 has joined #openttd
02:11:57  *** Wormnest has quit IRC
02:30:26  *** _aD has quit IRC
02:37:13  *** colde_ has joined #openttd
02:37:43  *** reldred has quit IRC
02:38:02  *** gregdek has quit IRC
02:39:08  *** mindlesstux has quit IRC
02:39:47  *** innocenat__ has joined #openttd
02:39:53  *** colde has quit IRC
02:39:53  *** colde_ is now known as colde
02:40:17  *** innocenat_ has quit IRC
02:40:17  *** innocenat__ is now known as innocenat_
02:40:40  *** twpol has quit IRC
02:45:35  *** mindlesstux has joined #openttd
02:55:44  *** D-HUND has joined #openttd
02:58:15  *** reldred has joined #openttd
02:59:08  *** debdog has quit IRC
02:59:39  *** twpol has joined #openttd
03:03:47  *** gregdek has joined #openttd
03:07:43  *** glx has quit IRC
03:44:22  *** Gustavo6046 has quit IRC
03:44:37  *** Gustavo6046 has joined #openttd
03:45:24  *** Flygon has joined #openttd
04:18:37  *** Gustavo6046 has quit IRC
04:19:26  *** Gustavo6046 has joined #openttd
04:39:27  *** Gustavo6046 has quit IRC
04:39:43  *** Gustavo6046 has joined #openttd
05:56:58  *** tokai|noir has joined #openttd
05:56:58  *** ChanServ sets mode: +v tokai|noir
06:04:00  *** tokai has quit IRC
06:44:04  *** andythenorth has joined #openttd
07:28:28  *** HerzogDeXtEr has joined #openttd
07:36:15  *** sla_ro|master has joined #openttd
07:40:43  * andythenorth wonders if PBS could be changed :P
07:40:52  <andythenorth> so signals aren't default red
07:41:08  *** Gustavo6046_ has joined #openttd
07:41:46  *** Gustavo6046 has quit IRC
07:41:46  *** Gustavo6046_ is now known as Gustavo6046
07:42:07  <andythenorth> oh
07:42:10  <andythenorth> I should just use block?
07:42:17  <andythenorth> is that why block is recommended practice?
07:42:46  *** jottyfan has joined #openttd
07:56:04  *** nielsm has joined #openttd
08:27:07  <DorpsGek> [OpenTTD/bananas-api] James103 commented on issue #24: Restrict dependencies between content-type
09:13:16  *** Gustavo6046 has quit IRC
09:13:31  *** Gustavo6046 has joined #openttd
09:41:21  *** Samu has joined #openttd
09:49:46  *** andythenorth has quit IRC
10:27:45  *** WormnestAndroid has quit IRC
10:33:17  *** andythenorth has joined #openttd
10:46:34  *** WormnestAndroid has joined #openttd
10:52:16  <FLHerne> andythenorth: It is in my book
10:52:21  <FLHerne> among other visual things
10:54:26  *** Samu has quit IRC
11:13:22  *** Samu has joined #openttd
11:26:19  <andythenorth> hmm
11:26:21  <andythenorth> new signals
11:26:23  <andythenorth> moar signals
11:26:31  <andythenorth> PBS, but block
11:26:44  <andythenorth> PBS that can be reserved through until next junction or red
11:27:47  *** andythenorth has quit IRC
11:49:15  *** Speeder has quit IRC
11:50:23  *** andythenorth has joined #openttd
11:58:26  *** andythenorth has quit IRC
12:08:34  *** glx has joined #openttd
12:08:34  *** ChanServ sets mode: +v glx
12:09:17  *** andythenorth has joined #openttd
12:09:21  <andythenorth> ha do we even know how PBS works? 😊
12:09:24  <andythenorth> or is it magic?
12:09:55  <DorpsGek> [OpenTTD/team] glx22 commented on issue #259: [ru_RU] Translator access request
12:29:07  *** WormnestAndroid has quit IRC
12:31:04  <glx> ok the funny part is stuck ship is not lost
12:31:21  *** jottyfan has quit IRC
12:32:42  *** Samu_ has joined #openttd
12:38:30  *** Samu has quit IRC
12:39:30  <glx> oh it's not lost because it doesn't even try to pathfind
13:05:52  *** virtualrandomnumber has joined #openttd
13:06:38  *** virtualrandomnumber has quit IRC
13:11:23  *** LouWestin has quit IRC
13:12:44  *** LouWestin has joined #openttd
13:31:07  *** WormnestAndroid has joined #openttd
13:40:28  *** Wormnest has joined #openttd
13:43:04  *** frosch123 has joined #openttd
13:47:54  *** sla_ro|master has quit IRC
13:58:03  <andythenorth> quak?
14:08:51  <Samu_> hi
14:15:38  <frosch123> moo
14:38:42  *** Gustavo6046_ has joined #openttd
14:39:20  *** Gustavo6046 has quit IRC
14:39:20  *** Gustavo6046_ is now known as Gustavo6046
14:46:27  *** sla_ro|master has joined #openttd
15:14:57  *** Samu_ has quit IRC
15:20:24  *** Beer has joined #openttd
15:38:17  *** Samu has joined #openttd
16:10:01  *** Gustavo6046 has quit IRC
16:14:23  *** Etua has joined #openttd
16:23:42  *** Gustavo6046 has joined #openttd
16:49:50  <glx> ok I think is the best solution
16:50:41  <glx> (includes the alternate version for #9567)
17:06:00  *** Flygon has quit IRC
17:09:30  <andythenorth> debugging a set of industries all writing to the same town storage
17:09:32  <andythenorth> is not fun
17:09:59  <glx> add debug lines in openttd
17:11:24  <andythenorth> that would require me to learn how to be a programmer :)
17:11:28  <andythenorth> oh dear :)
17:11:47  <andythenorth> also I'm caching a calculated value, which means I have 2 problems
17:12:05  <andythenorth> but I need to cache it so other grfs can read it without needing to encode the calculation
17:12:24  <andythenorth> even though it's a derivative of values in two other registers :)
17:12:34  * andythenorth learns about cache invalidation
17:12:51  *** Etua has quit IRC
17:14:31  <andythenorth> this only runs monthly, and I've nerfed ffwd performance by somehow compiling a debug build of openttd :D
17:14:36  <andythenorth> 'such andythenorth'
17:16:25  <andythenorth> is '-DCMAKE_BUILD_TYPE=RelWithDebInfo' what I need?
17:17:27  *** D-HUND is now known as debdog
17:19:09  <Timberwolf> Yes, the "fast, but prints stuff" option.
17:19:56  <andythenorth> thanks
17:20:06  <Timberwolf> Or rather, when it crashes you get a useful stack dump.
17:20:12  <andythenorth> :)
17:20:30  <andythenorth> hmm what if I just repaint PBS signal sprites to show green by default, not red?
17:20:36  <andythenorth> or maybe yellow by default?
17:21:04  <andythenorth> I can't unsee how the 'PBS is always red' thing now :)
17:21:28  <Timberwolf> I've got used to it.
17:22:09  <andythenorth> today was first time I noticed it :)
17:22:14  <Timberwolf> I think it's vaguely realistic?
17:23:06  <Timberwolf> Pretty sure I've used the "platform exit signal has gone from red to green" as early warning a Tube train is approaching.
17:23:52  <Timberwolf> Of course Tube signallimg != National Rail signalling.
17:26:41  <andythenorth> I think it depends on blocks vs route setting
17:26:44  <andythenorth> but that's a pure guess
17:26:56  <andythenorth> track circuit blocks will tend to green if not occupied
17:27:04  <andythenorth> if only we had...block signals :P
17:29:28  *** kaomoneus has joined #openttd
17:31:30  <kaomoneus> Hi fold! I'm trying to understand whe commands from commands.cpp are called? Week ago nielsm hinted me, that it is used for network interaction. But I can't find how it is called on client side..
17:32:44  <kaomoneus> E.g. I have CmdSetTimetableStart, and I want to change "day" with "hour". I can't fix all possible custom scripts, but at least I would like to fix network. But I can't find a client call for it.
17:33:22  <kaomoneus> .. I meant "hi folks" :-)
17:35:02  <frosch123> <- read that
17:38:31  <andythenorth> not sure I understand the results of industry_town_count
17:38:39  <glx> Timberwolf> Or rather, when it crashes you get a useful stack dump. <-- useful stackdump and apple don't work very well
17:39:19  <andythenorth> var 67 with bit 8 set
17:39:28  <andythenorth> it seems to under-count by one sometimes
17:39:58  <andythenorth> I wondered if it discounts 'self' in the case where queried industry ID is same as ID of caller
17:41:54  <andythenorth> the lack of a monthly town cb for grf :)
17:42:00  <andythenorth> means I have to fake it from industries
17:42:09  <andythenorth> which means counting how many industries there in the town
17:42:41  <andythenorth> and then having the last one called do the monthly town stuff
17:43:12  <andythenorth> maybe I should just copy the production code into GS?
17:43:51  <andythenorth> GS can run a monthly callback fine
17:44:01  <andythenorth> 'not a callback andythenorth'
17:44:13  <andythenorth> monthly event loop walking all towns
17:44:19  <kaomoneus> frosch123: ok, thanks, now I see.
17:49:19  <andythenorth> var 67 'probably fine'
17:49:24  <andythenorth> just add 1 every time :P
17:57:37  <andythenorth> lol I tried a monthly random industry production increase
17:57:46  <andythenorth> it works on a per-town basis
17:57:49  <andythenorth> but
17:58:06  <andythenorth> order industries are handled is deterministic :D
17:58:11  <andythenorth> so only potash mines are increasing
17:58:36  <andythenorth> (there's a flag to limit increases to one in any month)
18:01:26  *** sla_ro|master has quit IRC
18:01:31  *** kaomoneus has quit IRC
18:03:19  *** sla_ro|master has joined #openttd
18:05:07  *** gelignite has joined #openttd
18:18:32  *** kaomoneus has joined #openttd
18:22:16  <andythenorth> oof maybe I can patch GS to read the internal state of industries
18:22:24  <andythenorth> lol, probably beyond my skills
18:22:28  *** Beer has quit IRC
18:28:36  <andythenorth> hmm
18:28:48  <andythenorth> random production change cb anyone? :P
18:28:54  <andythenorth> is that broken with smooth economy?
18:29:21  *** gelignite has quit IRC
18:35:47  *** Wormnest_ has joined #openttd
18:37:50  *** WormnestAndroid has quit IRC
18:41:18  *** Wormnest has quit IRC
18:44:49  *** WormnestAndroid has joined #openttd
18:51:12  <andythenorth> could newgrf debug windows refresh when dirty?
18:51:29  <andythenorth> currently they have to be open/closed or windowshaded to update register values reliably
19:02:01  *** Wormnest_ has quit IRC
19:06:39  *** gelignite has joined #openttd
19:12:59  *** kaomoneus has quit IRC
19:15:26  *** Samu has quit IRC
19:26:27  *** WormnestAndroid has quit IRC
19:26:43  *** WormnestAndroid has joined #openttd
19:32:32  *** Samu has joined #openttd
19:36:02  <andythenorth> ouch really can't get a monthly town loop working
19:36:11  <andythenorth> maybe I do have to reimplement FIRS in GS
19:36:16  <andythenorth> this will be...weird
19:44:48  <andythenorth> hmm
19:44:49  <andythenorth> LOAD_PERM(${get_perm_num("this_cycle_industry_counter")}) < count_industries_in_town()
19:45:03  <andythenorth> the values for that are 18 and 9
19:45:08  <andythenorth> nml thinks 18 < 9
19:46:08  <glx> nml can't think, unless they are compile time constants
19:47:24  <andythenorth> I should be able to evaluate a storage and procedure result though? :)
19:48:03  <andythenorth>
19:48:57  <andythenorth> this just keeps counting up, it should end at 9 in my test case which is the current value of count_industries_in_town() :)
19:56:56  <glx> nml seems right, what's the corresponding nfo ?
20:01:39  * andythenorth looks
20:01:49  <andythenorth> I'm sure the bug is in my logic
20:01:58  <andythenorth> but I've stared at it for an hour or so
20:02:09  <andythenorth> this isn't the actual code, this is the attempt to debug
20:02:18  <andythenorth> but the debug code also needs debugging :P
20:03:41  <andythenorth> glx
20:21:02  <glx> except some weird r 1A 20 \dxFFFFFFFF sto 1A 20 \dx00000100, nfo seems correct
20:26:14  <andythenorth> hmm
20:27:05  <andythenorth> newgrf monthly cb is synchronous?  There's no chance of industries writing to storage whilst others are trying to read it?
20:27:17  <andythenorth> openttd code looks like a simple iterator to me
20:27:30  <glx> everything is sequential in openttd world
20:27:38  <glx> because determinism
20:30:21  <andythenorth> this is only called once per industry
20:30:22  <andythenorth> hmm
20:30:25  * andythenorth tries more things
20:31:36  <glx> but I still don't get why there are temp[256]=-1 in the middle of ?: nfo
20:34:26  <frosch123> that is for the town-storage access
20:34:29  <andythenorth> emoji
20:34:34  <frosch123> "-1" is for "this grf"
20:34:40  <glx> oh ok
20:39:25  <andythenorth> ha ha
20:39:38  <andythenorth> I have to compile once with a version to put storage back to 0 in my savegame
20:39:47  <andythenorth> then compile again with the actual code
20:39:49  <andythenorth> this is lol :)
20:40:05  <glx> store_perm translation feels weird in this case temp[256]=-1; temp[256]=-1; load_perm[1]; add 1; store_perm[1]=res
20:40:31  * andythenorth saves a savepoint
20:42:37  <andythenorth> I have simplified my conditional to `LOAD_PERM(${get_perm_num("this_cycle_industry_counter")}) < 9`
20:42:59  <andythenorth> to avoid the procedure call
20:43:23  <andythenorth> but still evaluating 'true' for 12 < 9'
20:49:20  <andythenorth> this gets me 101 in the register
20:49:42  <andythenorth> which suggests that 100 < 9
20:50:48  *** Gustavo6046_ has joined #openttd
20:52:07  *** Gustavo6046 has quit IRC
20:52:37  <glx> but nfo looks correct
20:54:08  <glx> cmp(100, 9) should be 2, transformed to 1 by min(2,1), then transformed to 0 by xor 1
20:54:32  <andythenorth> hmm
20:54:44  <andythenorth> scope here is town
20:54:48  <nielsm> reading this conversation gives me two thoughts: first, it'd (still) be good to have NML write a map file, giving a two-way mapping from NML source lines to GRF sprite numbers
20:55:13  <andythenorth> maybe I should test with scope SELF
20:55:37  <nielsm> and second, it would probably be useful to have a way to at least get detailed dumps of callback evaluation
20:55:40  *** Wormnest has joined #openttd
20:55:57  *** Gustavo6046_ is now known as Gustavo6046
20:56:20  <nielsm> disadvantage of the latter: stuffing more code into the callback evaluation, that would be skipped most of the time, probably slowing everything down all the time
20:56:22  <glx> oh I know what happens, it's a side effect
20:56:33  *** WormnestAndroid has quit IRC
20:57:09  <frosch123> nielsm: <- the backtrace code is not that bad
20:57:15  <frosch123> but the gui is too complicated
20:57:30  <glx> of (!guard*0)+(guard*store_perm)
20:57:34  <nielsm> "4712 commits behind OpenTTD:master" ouch
20:57:48  <glx> with guard being 0 here
20:57:56  *** Samu has quit IRC
20:58:00  <glx> but the store_perm is still executed
20:58:14  <frosch123> nielsm:
20:58:35  <andythenorth> o_O
20:58:35  *** Gustavo6046 has quit IRC
20:58:54  <glx> andythenorth: I think you should rewrite it to not use ?:
20:59:25  <frosch123> the premium solution would be a rest interface to ottd, so some external program could trigger callbacks, provide the GUI, visualize stuff, and read map files from NML
20:59:26  <glx> keep the < in expression, but the decision with 0: and 1:
20:59:43  <glx> or just 1: and default
20:59:54  <frosch123> nielsm: my master is not upstream master :p
21:00:01  <frosch123> it may be thousands more revisions behind :p
21:00:45  *** WormnestAndroid has joined #openttd
21:01:02  <andythenorth> I can't put an expression into a result, I have to chain to another switch or return a constant?
21:01:10  * andythenorth never really understood nml
21:01:16  <nielsm> yep that branch looks like it's on top of not-quite-1.4
21:01:35  <glx> you can put an expression, nml will generate the switch if needed
21:01:36  <andythenorth> e.g. can't do `1: STORE_PERM(100, 230);` or similar
21:01:41  <andythenorth> oh I can?
21:02:27  * andythenorth wonders if `return` is mostly superfluous
21:03:02  <glx> maybe nml could detect if ?: may have side effects and at least warn about it
21:03:46  *** Gustavo6046 has joined #openttd
21:04:07  <glx> I think detection is possible thanks to is_read_only
21:04:26  <andythenorth> :)
21:05:01  <andythenorth> nmlc seems quite happy with this
21:05:08  * andythenorth has never tried this before
21:05:18  <andythenorth> I guess it just expands a switch then
21:05:41  <glx> return is useless here :)
21:06:56  <andythenorth> I should learn nml
21:09:56  *** Samu has joined #openttd
21:10:13  <andythenorth> ok so that's the debugging code fixed :)
21:10:27  <andythenorth> now I just need to find out why the actual code doesn't work
21:13:49  <glx> and I think I spotted a small bug in is_read_only() for ternary
21:13:59  <glx> it doesn't check guard
21:17:13  *** Gustavo6046_ has joined #openttd
21:19:32  <andythenorth> hmm
21:20:16  *** _aD has joined #openttd
21:20:50  *** Gustavo6046 has quit IRC
21:20:50  *** Gustavo6046_ is now known as Gustavo6046
21:25:46  <andythenorth> I would very like to be able to inspect a backtrace right now :P
21:27:01  <Samu> just found a miniscule bug in my airport building code
21:27:01  *** Gustavo6046 has quit IRC
21:28:13  *** frosch123 has quit IRC
21:28:58  <Samu> basically, if it finds no space to build an airport of type A, it was skipping testing the next airport type
21:29:17  <Samu> it should be testing all types
21:29:24  <Samu> it's fixed now :)
21:30:26  <Samu> on a map with huge towns and tiny islands, this was happening quite often, nearly no airports were being built, because the first thing it tests is intercontinental
21:30:50  <Samu> and the ai is able to build heliports
21:31:16  <Samu> but wasn't even trying
21:31:41  <Samu> gotta go sleep, take care
21:33:57  *** Samu has quit IRC
21:35:34  *** Gustavo6046 has joined #openttd
21:40:02  <andythenorth> hmm another odd result
21:41:14  <glx> andythenorth:
21:41:30  <andythenorth> :)
21:41:30  <glx> warning message may not be the best
21:42:02  <glx> but at least it should tell something won't do what you expect
21:43:09  <andythenorth> thanks :)
21:43:30  <glx> ternary implementation is weird and smart at the same time, because NFO limitations
21:44:06  <andythenorth> should we merge it :)
21:47:35  <andythenorth> ok got another weird result :)
21:48:47  <glx> and after looking at the nfo generated in your "broken" test case, I can imagine a potential issue for things like store_perm(procedure_setting_temp0x100(), reg_num)
21:49:04  <andythenorth> got a procedure called in an expression, doesn't seem to be called unless I add more things to the expression
21:49:10  <andythenorth> hard to explain, and code is large
21:49:16  * andythenorth trying to isolate the failure
21:51:36  <andythenorth> glx is that where 0x100 gets over-written?
21:51:48  <glx> yes
21:51:58  <andythenorth> there are similar cases with 0x101 elsewhere (parameter to multiple vars)
21:52:19  <andythenorth> can't remember the details but I encountered it before
21:55:24  <andythenorth>
21:56:11  <glx> lol even regression test 039 shows the issue
21:56:32  <andythenorth> ^ if I don't include the `STORE_TEMP` after call to monthly_town_loop(), the procedure call appears to fail
21:57:02  <glx> and
21:59:15  <glx> fail or optimised ?
21:59:53  <glx> IIRC procedure calls should not be optimised in case of constant result
22:01:56  <glx> hmm seems it's the default result only case, so yeah probably optimised
22:02:06  <glx> you disabled the warnings ?
22:03:19  <andythenorth> yes, there are a lot :)
22:03:39  <glx> if the procedure is read only, it will be skipped
22:03:53  <glx> because you return a fixed constant
22:03:56  <andythenorth> ok
22:04:15  <andythenorth> so I just work around that in this case :)
22:04:55  <andythenorth> ok that explains why `monthly_town_loop(), 1` makes no difference
22:05:03  <andythenorth> but a store temp does
22:05:36  <glx> if monthly_town_loop() only reads stuff it's useless
22:06:17  <andythenorth> well it does store perm
22:06:30  <andythenorth> but it's several layers of procedure
22:06:44  <glx> hmm in this case it may be a bug somewhere
22:07:07  <andythenorth> :)
22:10:11  <glx> <-- I think it's this optimisation, but is_read_only() should recurse
22:10:49  * andythenorth bug-finding-as-a-service :)
22:10:51  <glx>
22:11:00  <glx> maybe there's a bug here
22:13:08  <glx> I guess it should probably do something similar to collect_references
22:17:24  <andythenorth> :)
22:18:42  <andythenorth> ha well now I know what the issue is :)
22:18:50  <andythenorth> I've worked around it and can test more code
22:18:51  <andythenorth> thanks
22:20:15  <glx> I added a possible solution to your gist
22:20:38  <andythenorth> cool
22:20:44  <glx> not having a test case locally
22:22:23  * andythenorth tests
22:24:11  <andythenorth> maybe `False` not `false` :)
22:24:41  <glx> hey regression pass here :)
22:25:11  <glx> because I don't have a valid test case which could enter in this branch
22:25:56  <glx> but that means it detected a non read only thing
22:26:12  <andythenorth> seems to work
22:33:08  *** Gustavo6046_ has joined #openttd
22:34:56  <andythenorth> merge? :) :P
22:35:37  *** Gustavo6046 has quit IRC
22:35:37  *** Gustavo6046_ is now known as Gustavo6046
22:40:54  * andythenorth must to sleep
22:41:12  <glx> I think I need to add some test cases
22:41:22  <andythenorth> thanks for the help glx
22:41:32  *** andythenorth has quit IRC
22:41:35  *** sla_ro|master has quit IRC
22:53:38  *** Gustavo6046_ has joined #openttd
22:57:00  *** Tirili has joined #openttd
22:57:49  *** Gustavo6046 has quit IRC
23:03:38  *** Gustavo6046_ has quit IRC
23:09:47  *** nielsm has quit IRC
23:10:38  *** Gustavo6046 has joined #openttd
23:20:35  *** WormnestAndroid has quit IRC
23:20:36  *** WormnestAndroid has joined #openttd
23:32:43  *** WormnestAndroid has quit IRC
23:33:14  *** WormnestAndroid has joined #openttd

Powered by YARRSTE version: svn-trunk