Config
Log for #openttd on 29th January 2023:
Times are UTC Toggle Colours
10:23:55  *** sla_ro|master has joined #openttd
10:45:37  <andythenorth[d]> trying to replicate
10:46:08  <andythenorth[d]> I haven't been able to trigger it yet
10:47:09  <nielsm> it sounds like you'd have to share at least your config files, and possibly your entire collection of installed content
10:49:33  *** reldred has quit IRC
10:50:55  <Wolf01> https://pastebin.com/9yuS23dV
10:54:21  <andythenorth[d]> does GS get any free opcodes when loading savegame?
10:55:52  <andythenorth[d]> I am trying to make my GS as stateless as possible with respect to savegame, but that requires doing slow things at least once
10:58:56  <Wolf01> https://pastebin.com/4DCu1EXJ and the downloaded content
11:20:17  <nielsm> andythenorth[d]: the "load saved data" function is kind-of free opcodes, but I don't think you're allowed to touch game state during it, only calculate things internally (can't fire commands)
11:20:56  *** Samu has joined #openttd
11:22:51  <andythenorth[d]> thanks
11:22:58  <andythenorth[d]> but I might be able to do stuff like walk all the tiles on the map
11:29:07  <nielsm> I'd suggest (again) to try to save as much of your internal state as possible in a data structure that can be loaded and saved directly by the load/save functions
11:29:27  <nielsm> so you can just load it and be running straight away without having to re-calculate anything
11:36:01  <Samu> got a weird crash
11:36:22  <Samu> i opened Industry chains, boom
11:39:31  <Samu> let me test on master
11:39:54  <andythenorth[d]> I want to drop most of the state on saveload 🙂
11:40:08  <andythenorth[d]> it will be a source of bugs
11:40:26  <andythenorth[d]> and will require migrations, or constantly building a new test game
11:41:35  <andythenorth[d]> specifically I am copying a lot of the map into a squirrel table
11:42:21  <Xarick> https://cdn.discordapp.com/attachments/1008473233844097104/1069220957971034123/image.png
11:42:21  <Xarick> current master
11:50:55  <DorpsGek> [OpenTTD/OpenTTD] SamuXarick opened issue #10430: [Crash]: Opening Industry chains window https://github.com/OpenTTD/OpenTTD/issues/10430
12:04:30  <petern> Can you try in RC2?
12:19:38  <Samu> does not crash in rc2
12:20:02  <petern> Excellent 🙂
12:25:53  <andythenorth[d]> hmm my GS needs to stop deleting all of the map 😛
12:25:54  <andythenorth[d]> oops
12:26:02  <andythenorth[d]> I left it in as a test
12:26:21  <andythenorth[d]> but in normal gameplay it takes about 5 years to do it 😛
12:26:28  <andythenorth[d]> such savegame
12:26:39  <andythenorth[d]> GS lolz
13:32:02  <TrueBrain> I am looking (once again in sorts) if we could upload crashes automatically for release-builds .. it seems pretty doable, the main issue is: most solutions don't ask the user to report the crash, but just upload it
13:32:38  <TrueBrain> the days of "Report this issue" seems to be somewhat over .. and most software just sends it to a central server
13:33:49  <TrueBrain> there also seems to be no real cross-platform way to add a `.log` file to crashes .. but looking into how the crash-collectors work, that is not really surprising
13:36:50  <TallTyler> Somewhat related because that sounds like a lot more bug reports would be generated, I wonder if a permanent flag like “money cheat has been used in this savegame” but instead for “NewGRF has been changed in a running game” would be a useful addition to savegames, so we could more quickly triage those. Although I guess most of those reports are “economy broke plz fix” and not “the game crashed”. I guess I’m just
13:37:16  <TrueBrain> changing NewGRFs are tracked by the game log
13:37:30  <TrueBrain> you can see that in the `crash.log` currently
13:37:44  <TrueBrain> (game logs are embedded in the savegame)
13:41:22  <TrueBrain> take for example this crashlog: https://github.com/OpenTTD/OpenTTD/issues/10108#issuecomment-1291134247
13:41:30  <TrueBrain> there you can see that on tick 1280, NewGRFs were changed
13:46:59  <TallTyler> Ah, guess I never learned to read crash logs properly then 😛
13:47:20  <TrueBrain> it also doesn't happen all that often, that a crash is reported due to players doing silly things with NewGRFs
13:47:44  <TrueBrain> but sometimes players do, and we love to point out: you ignored a bit-fat-warning, so .. yeah ... 😛
13:49:55  <TrueBrain> owh, for crashpad, the most likely backend for Windows / MacOS, we can indicate it should also attach a log-file on crash .. so an option would be to write the gamelog to disk, and append when things change, so it gets attached on crash
13:53:12  <andythenorth[d]> We’d have opt out? And we could black some things?
13:54:04  <andythenorth[d]> E.g I always redact crashlogs because the mac ones contain 50% of my auth creds in the path
13:54:30  <TrueBrain> There is a tip .. don't use your username as part of your password 😛
13:55:01  <andythenorth[d]> You know what I mean :p
14:00:10  <TrueBrain> okay, so this would also drag in libcurl, but otherwise seems doable
14:00:35  <TrueBrain> would need to use vcpkg for Linux I guess, to keep things sane .. as crashpad is not in the Ubuntu package repo 😛
14:11:51  *** D-HUND has joined #openttd
14:14:27  <TrueBrain> okay, so there is sentry-native, which builds against all our targets. Crashpad as backend. Sentry Relay in `openttd.org`, so there are no weird calls to weird domains from the client. Opt-in on first-start to ask: automatically upload crashes? vcpkg for sentry-native and libcurl .. and that is all she wrote .. guess worth a test
14:15:47  <TrueBrain> it would upload a minidump and the gamelog of the active game. By uploading the debug-symbols to sentry, this should give pretty accurate stacktraces for crashes.
14:16:03  *** D-HUND is now known as debdog
14:16:35  <TrueBrain> and of course we should only enable it on our releases, as you don't want to receive this bla on development builds 😄
14:17:22  <TrueBrain> it would however be incompatible with our current crash handlers .. that is the only real downside I can find. So no more `crash.sav`, basically
14:17:56  <TrueBrain> makes me wonder .. was that ever actually useful? Can't remember 😛
14:24:37  <LordAro> sometimes
14:24:43  <LordAro> depends on the crash
14:26:04  <TrueBrain> well, we would lose that ability. Similar for the `crash.png` file
14:34:12  *** dude[m] has quit IRC
15:15:52  <kamnet> If the new way makes it easier and faster to diagnose and fix issues with less waiting on the public to report it, I don't see it as a loss.
15:16:09  <TrueBrain> yup
15:16:25  <TrueBrain> there is a balance ... we need to estimate how much reports we would receive, as storage is not endless 😛
15:17:42  <TrueBrain> but luckily enough sentry de-duplicates similar stacktraces, and doesn't keep all the minidumps around 😛
15:26:42  *** Wormnest has joined #openttd
16:03:33  <DorpsGek> [OpenTTD/OpenTTD] rubidium42 opened pull request #10431: Fix #10430, Fix 6ba55e6: display chain window causing assert https://github.com/OpenTTD/OpenTTD/pull/10431
16:24:41  <Samu> is github suddenly slow, or is it me?
16:50:36  *** Flygon has quit IRC
17:11:42  <DorpsGek> [OpenTTD/OpenTTD] rubidium42 merged pull request #10431: Fix #10430, Fix 6ba55e6: display chain window causing assert https://github.com/OpenTTD/OpenTTD/pull/10431
17:11:45  <DorpsGek> [OpenTTD/OpenTTD] rubidium42 cl

Powered by YARRSTE version: svn-trunk