| 19 Sep 2024 |
 Yorusaka Miyabi | In reply to @shadowrz:nixos.dev
It seemd to work, but then UIAA flow broke due to https://matrix.nixos.dev/_matrix/client/r0/auth/m.login.sso/fallback/web failed with M_UNRECOGNIZED
Also related that https://matrix.nixos.dev/_matrix/client/v3/auth/m.login.sso/fallback/web also errors with "Unrecognized Request" | 23:00:59 |
 Andreas Rammhold | interesting, I thought I handled all the SSO URLs. Actually tried an entire login-in | 23:14:10 |
| Andreas Rammhold | The worker documentation is really lacking or written in a way that makes me trip over every single pitfall. there is /_matrix/client/$version/devices prefix but obviously the delete_device isn't underneath but its own thing. | 23:18:18 |
| Andreas Rammhold | Loggin in works for me. 夜坂雅 let me know if it works for you. Perhaps that is a client specific thing depending on the flow it is picking? | 23:19:47 |
| Yorusaka Miyabi | In reply to @andir:nixos.dev
Loggin in works for me. 夜坂雅 let me know if it works for you. Perhaps that is a client specific thing depending on the flow it is picking? Login works, the https://matrix.nixos.dev/_matrix/client/v3/auth/m.login.sso/fallback/web is a UIAA fallback endpoint defined in the spec so clients can use a browser to complete re-authorization, UIAA is used for some operations like logout other devices | 23:21:29 |
| Andreas Rammhold | Did that ever work? | 23:21:52 |
| Andreas Rammhold | It is returning a missing session thing now so that looks promising? | 23:22:28 |
| Yorusaka Miyabi | In reply to @andir:nixos.dev
It is returning a missing session thing now so that looks promising? Yes, and the re-authorization flow also works | 23:23:08 |
| Yorusaka Miyabi | In reply to @andir:nixos.dev
It is returning a missing session thing now so that looks promising? * | 23:23:14 |
| Andreas Rammhold | Great! Thank you for all the feedback and sorry for breaking it :) | 23:23:21 |
| 20 Sep 2024 |
| Yorusaka Miyabi | In reply to@andir:nixos.dev
Great! Thank you for all the feedback and sorry for breaking it :) Oh and also https://matrix.nixos.dev/_matrix/client/v3/keys/signatures/upload broke too which makes Cinny insists my new session is unverified | 08:32:41 |
| Yorusaka Miyabi | In reply to@shadowrz:nixos.dev
Oh and also https://matrix.nixos.dev/_matrix/client/v3/keys/signatures/upload broke too which makes Cinny insists my new session is unverified (And it can't be properly verified by Cinny) | 08:38:30 |
| Yorusaka Miyabi | In reply to@shadowrz:nixos.dev
Oh and also https://matrix.nixos.dev/_matrix/client/v3/keys/signatures/upload broke too which makes Cinny insists my new session is unverified OK it seems to work now | 08:39:12 |
 andi- | I just deployed a special case for the keys url | 08:39:21 |
| Yorusaka Miyabi | In reply to@andi:kack.it
I just deployed a special case for the keys url Oh. | 08:39:41 |
| andi- | https://element-hq.github.io/synapse/latest/workers.html#synapseappgeneric_worker it is funny how this stuff is so broken | 08:39:50 |
| andi- | there is a list of all the URLs the generic workers are supposed to support | 08:40:02 |
| andi- | yet I've to create plenty of execeptions because it simply doesn't work | 08:40:16 |
| andi- |
These endpoints can be routed to any worker.
Yet, only the main process does it for us.
| 08:40:57 |
| andi- | I also can't find many/any bug reports for this... Perhaps it is me holding it wrong? I'll have another read on the doc... I was actually happy how smooth it was before these kinds of errors surfaced. | 08:47:23 |
| Yorusaka Miyabi | In reply to@andi:kack.it
I also can't find many/any bug reports for this... Perhaps it is me holding it wrong? I'll have another read on the doc... I was actually happy how smooth it was before these kinds of errors surfaced. Just asked in https://matrix.to/#/#synapse:matrix.org they suggested to provide reverse proxy config to troubleshoot | 09:03:12 |
| Yorusaka Miyabi | (You can join too I guess) | 09:03:23 |
| Yorusaka Miyabi | In reply to@shadowrz:nixos.dev
Just asked in https://matrix.to/#/#synapse:matrix.org they suggested to provide reverse proxy config to troubleshoot (And also related worker config) | 09:04:43 |
| andi- | It is still syncing :D | 09:13:00 |
| Andreas Rammhold | 夜坂雅: did roll back the sendToDevice, read markers, presence and push-rules, account_data, ... overrides. Lets see if this still works. I'll keep the SSO on the main process for now. Given the limited amount of logins we have that isn't a concern right now. | 09:53:24 |
| andi- | doing the reverse approach now, I'll only redirect URLs from the document to the workers... that seems like the smaller set of URLs or less work long-term :( | 10:21:43 |
| Yorusaka Miyabi | In reply to @andir:nixos.dev
夜坂雅: did roll back the sendToDevice, read markers, presence and push-rules, account_data, ... overrides. Lets see if this still works. I'll keep the SSO on the main process for now. Given the limited amount of logins we have that isn't a concern right now. Mostly works,
[...] the left remaining API requests that failed with M_UNRECOGNIZED on nixos.dev are the pushers API and /_matrix/client/v3/directory/list/room/ (which deals with whether a room is public on the server's public directory), but they're otherwise not important to fix and can be delayed.
(Pushers API goes to the main process)
| 10:22:04 |
| Andreas Rammhold | This is now live. Lets see if that remove the need for random overrides that aren't really following a pattern / have a documentation entry. | 10:22:50 |
| Andreas Rammhold | test | 10:25:53 |
| andi- | Funny part is that the urls for the worker in the doc also include URLs that need to be routed to the respective stream writer... There must be a better way to document that. | 10:24:56 |
