| 14 Jun 2022 |
FantasyCookie17 🏳️🌈🏳️⚧️ | Why else would it think it's offline when it's not? | 14:58:30 |
Dandellion | its supposed to be offline | 14:58:50 |
Dandellion | there's a SIGKILL in your log which looks a little sus | 14:58:56 |
FantasyCookie17 🏳️🌈🏳️⚧️ | Hm… where could that be coming from? | 14:59:39 |
Dandellion | did you try building it more times? | 15:01:25 |
Dandellion | are you sure you didnt hit ctrl+c upon seeing the warnings or something? | 15:01:39 |
FantasyCookie17 🏳️🌈🏳️⚧️ | Yes… several times, and no Ctrl C… | 15:01:54 |
Dandellion | huh, then that is strange | 15:02:01 |
FantasyCookie17 🏳️🌈🏳️⚧️ | Ctrl+C would be SIGTERM, not SIGKILL, anyway, afaik… | 15:02:37 |
Dandellion | yaeh sigint | 15:03:08 |
Dandellion | * yeah sigint | 15:03:12 |
FantasyCookie17 🏳️🌈🏳️⚧️ | https://github.com/jupyterhub/jupyterlab-hub/issues/44
It… could be a memory issue? | 15:08:44 |
FantasyCookie17 🏳️🌈🏳️⚧️ | But our machine has 4GB… | 15:08:54 |
FantasyCookie17 🏳️🌈🏳️⚧️ | Maybe… if we add a swapfile… | 15:09:08 |
FantasyCookie17 🏳️🌈🏳️⚧️ | That… seems to have fixed it… | 16:36:13 |
Ralith | 4G isn't much these days | 19:43:27 |
Ralith | especially when there's web crap involved | 19:44:02 |
| 16 Jun 2022 |
FantasyCookie17 🏳️🌈🏳️⚧️ | It's an ARM machine… and it's more than enough for what it usually runs (Caddy + WireGuard + BorgBackup twice a day)… | 04:44:47 |
@linus:schreibt.jetzt | Builds often do need more memory though | 08:34:50 |
| * @linus:schreibt.jetzt has flashbacks to ARMv7 system that couldn't compile chromium because it didn't have enough address space | 08:35:21 |
@linus:schreibt.jetzt | (Or was it Firefox?) | 08:35:45 |
| 17 Jun 2022 |
fadenb | error: The option `services.matrix-synapse.settings.listeners.[definition 1-entry 1].bind_address' does not exist. Definition values:
- In `/etc/nixos/configuration.nix': "127.0.0.1"
(use '--show-trace' to show detailed location information)
Did I misinterpret the latest release config that all top-level config things should be moved to the settings. section?
| 11:41:49 |
fadenb | Ah 🤦♂️, I should have read the next section 😆 | 11:42:22 |
hexa | tl;dr: bind_addresses | 16:12:40 |
hexa |
The listeners.*.bind_address option was renamed to bind_addresses in order to match the upstream homeserver.yaml option name. It is now also a list of strings instead of a string.
| 16:13:50 |
hexa | wow, I made that it it's own sentence 😄 | 16:14:03 |
fadenb | I ran into another issue though.
I naively just added a settings = { indented my old config lines and closed the curly braces.
This was valid enough to build but had multiple issuss:
database_args was accepted but did not work as expected. This lead to my database name below that being ignored.
Fix was easy, make it database.args and it is ok.
Another thing which broke was my old extraConfig which ended up as a single line in the new config.
I feel this was not sufficiently documented. For such a major change I'd hope for more safeguards or a much more detailed release notes entry that actually covers all those cases. | 19:20:39 |
Dandellion |
need to be moved into services.matrix-synapse.settings
might have kind of implied you could just copy it in. but how settings (RFC42) works in the rest of nixos should make it somewhat a pattern so people hopefully don't actually verbatimely copy things into the settings
I don't think there's a nice way to sanity check freeform settings in general (since you want to be able to specify their contents freely), but maybe we could have explicitly checked for things like database_args for a while and thrown
| 20:00:18 |
Dandellion | I personally think saying it generates a homeserver.yaml file and follows RFC42 ought to be enough, at least combined with settings being like that in other modules | 20:03:39 |
| 21 Jun 2022 |
fadenb | Anyone here using synapse_auto_compressor with psql socket auth and can share with me how to specify how to connect to postgres? | 21:15:31 |