IRC logs for #openttd on OFTC at 2020-08-20
⏴ go to previous day
00:20:34 *** WormnestAndroid has quit IRC
00:21:00 *** WormnestAndroid has joined #openttd
00:47:03 *** Wormnest has joined #openttd
00:53:31 *** WormnestAndroid has quit IRC
00:54:23 *** WormnestAndroid has joined #openttd
02:18:45 *** greeter has joined #openttd
02:37:25 *** WormnestAndroid has quit IRC
02:38:22 *** WormnestAndroid has joined #openttd
02:39:26 *** WormnestAndroid has quit IRC
02:39:40 *** WormnestAndroid has joined #openttd
08:09:56 *** sla_ro|master has joined #openttd
09:17:56 *** andythenorth has joined #openttd
10:27:00 <andythenorth> colour me surprised
10:28:21 *** arikover has joined #openttd
11:12:39 <TrueBrain> are you spying on me?
11:13:30 <TrueBrain> well, given you timed it at the same moment I started my client, that was a bit scary :(
11:14:25 <andythenorth> I should bet on some horses
11:14:32 <andythenorth> clearly my daily RNG seed is good
11:14:45 <andythenorth> I really do think we get a daily RNG in life
11:16:00 <TrueBrain> copying to an external disk with 100MB/s .. I have seen worse
11:23:24 *** Smedles_ has joined #openttd
13:32:25 <TrueBrain> why is the MSU protocol so idiotic :(
13:32:31 <TrueBrain> the deregister packet doesn't contain the server key :(
13:50:14 <TrueBrain> looking into DynamoDB, as it seems an obvious cheap solution for the MSU .. but boy, that needs another way of thinking :P
14:00:37 *** andythenorth has joined #openttd
14:43:45 *** WormnestAndroid has quit IRC
14:48:08 <DorpsGek_III> [OpenTTD/OpenTTD] James103 commented on issue #8296: Non-latin text language not showing in mac os https://git.io/JJxmQ
14:49:25 *** WormnestAndroid has joined #openttd
14:50:13 *** ChanServ sets mode: +v tokai
14:50:20 *** WormnestAndroid has joined #openttd
16:17:20 <TrueBrain> MSU is weird .. every 15 minutes, a server announces itself to the MSU
16:17:28 <TrueBrain> every 5 minutes, the MSU polls the server if it is still alive
16:17:34 <TrueBrain> I am somewhat tempted to drop the last part
16:18:00 <LordAro> does seem a bit redundant
16:18:02 <TrueBrain> I wonder what we lose .. well, except that servers can remain online for ~20 minutes after they go offline (5 minute grave period)
16:18:14 <TrueBrain> I think the idea behind it was to mark servers offline earlier than the broadcasts
16:19:04 <TrueBrain> but in the worst case, there are entries in the serverlist for 20 minutes that you cannot fetch information about .. I mean .. not the worst problem
16:20:06 <TrueBrain> dropping the "query" part would simplify the implementation a lot
16:20:39 <LordAro> could decrease the announce interval?
16:20:46 <TrueBrain> hardcoded in the client
16:20:59 <LordAro> client can be changed
16:21:00 <TrueBrain> so only for new versions we can do that :)
16:21:05 <TrueBrain> but .. does it matter?
16:21:07 <TrueBrain> 15 minutes sounds fine
16:21:36 <TrueBrain> I mean, literally the worst thing that would happen is: you announce yourself, your server crashes, and you stay in people's network list for 20 more minutes ... where the client cannot fetch even the servername
16:25:13 <TrueBrain> but in general, so many things are ... outdated, when it comes to the MSU :)
16:32:30 <TrueBrain> okay, the current MSU already removes your server if you stop advertising, after 20 minutes
16:33:02 <TrueBrain> so the updater really only gives better information about the server over time .. every 5 minutes instead of every 15 minutes :)
16:33:21 <TrueBrain> guess that for my replacement MSU, I am going to skip that part for now :)
16:34:10 <TrueBrain> (as it is rather complicated the implement .. 1 server can have multiple IPs, which are all queried for an update, most of the time close at the same time, where "offline" becomes a bit of a difficult statement .. if 1 IP doesn't respond, does that make the server offline? Etc .. lot of code currently taking care of all those edge-cases)
16:37:02 <TrueBrain> _dp_: you run servers .. do you have an opinion on this? :)
16:40:27 <_dp_> hi, I think rationale behind the query from msu was to check that server can actually be joined (i.e. port forwarding)
16:41:36 <_dp_> which seems totally reasonable thing to have
16:42:15 <_dp_> except I have a sneaking suspicion that it sometimes fail and servers disappear from the list but that's a slightly different story
16:54:13 <TrueBrain> yeah, but that already happens on registering :)
16:54:23 <TrueBrain> when you start the server, it first checks if the ports are available, before adding them
16:54:32 <TrueBrain> so we don't lose that functionality
16:54:44 <TrueBrain> added bonus, the scenario you describe won't happen again, as the server will simply retry :)
16:55:01 <TrueBrain> (well, the updater tries 3 times too, but the server never sees it when they don't succeed)
16:55:08 <TrueBrain> another reason not to do the updater
16:55:26 <TrueBrain> from what I can tell, the only loss really is the resolution of the info on servers.openttd.org
16:55:38 <TrueBrain> guess we can simply try to do it without, and see what breaks :P
16:56:04 <TrueBrain> URLs of servers.openttd.org are going to change anyway, so people scraping the HTML will have to redo their work :)
16:56:32 <TrueBrain> who maintains that website that shows graphs about how many people were online? :)
17:05:06 *** Wormnest has joined #openttd
17:13:14 *** Wolf01 is now known as Guest454
17:14:47 *** Wolf01 is now known as Guest455
17:15:27 <_dp_> well, one on citymania is obv me, so it's not a problem
17:15:37 <_dp_> one on novapolis is TheDude I'd guess
17:15:47 <_dp_> and there may be some else I don't remember
17:42:40 <TrueBrain> Well, you get a real API for it in return .. so hopefully that is easy to fix :D
17:43:32 <_dp_> yeah, api would be much better
17:43:37 <_dp_> though I was kind of thinking to do a server scan myself
17:44:19 <_dp_> actually, iirc novapolis does server scan so shouldn't be affected
18:04:01 *** Progman has joined #openttd
18:17:59 *** gelignite has joined #openttd
18:33:32 *** frosch123 has joined #openttd
19:01:02 <TrueBrain> okay, that heavily simplified the MSU. Got most of it working now .. no clue what it is going to do for the costs, but we have some room to find that out :)
19:01:13 *** andythenorth has joined #openttd
19:01:32 <TrueBrain> I see the light at the end of the tunnel now :D
19:02:12 <frosch123> maybe there is a fire
19:02:40 <TrueBrain> wow ... writing the inode table on this disk is taking for ever .. what is going on ..
19:06:26 <TrueBrain> okay, time to export some old VMs of our cluster and store it in cold storage :)
19:18:35 *** HerzogDeXtEr has joined #openttd
20:01:05 <frosch123> running an aggressive gc reduced the wiki repo size from 1GB to 600MB
20:01:16 <frosch123> "git log" is still slow though
20:17:58 <TrueBrain> I am not really surprised, I have to admit :P
20:18:13 <TrueBrain> how many files btw? (in HEAD)
20:21:40 <frosch123> also 239 pages in "Mediawiki talk" is ridiculous
20:22:09 <TrueBrain> BaNaNaS also got slow with 40k files
20:22:13 <TrueBrain> guess that is also part of the reason :)
20:22:50 <frosch123> well, it's more like i am using it in a weird way
20:23:31 <frosch123> "git log" takes 1.3s. "git log MainPage" takes 81s
20:23:46 <frosch123> apparently git has to extract all diffs to know whether a commit touches a file
20:25:58 <b_jonas> frosch123: is that because it tries to detect merges?
20:26:57 <frosch123> is that a question or an answer?
20:27:12 <frosch123> does not sound plausible to me.
20:27:28 <b_jonas> I don't understand git
20:27:56 <b_jonas> do you have any suspicious options in git config -l or git config --global -l ?
20:40:37 *** andythenorth has joined #openttd
23:01:53 <b_jonas> these local authorities value trees to be worth so much more than houses and roads.
23:13:28 *** andythenorth has joined #openttd
23:47:30 *** virtualrandomnumber has joined #openttd
23:48:21 *** virtualrandomnumber has quit IRC
continue to next day ⏵