penguin
penguin#5453
Joined Oct 6, 2026 @SolunaCG on GitHub
Sign in to search- On Monochrome icon not matching other icon sizes on Gnome DesktopStill happening on KDE Plasma with Canary Desktop 2026.1116.70910 (Fedora 45, x64). Since the latest update, the monochrome Fluxer tray icon is significantly larger than the other system tray icons. The icon was previously displayed at the correct size after the fix for #4042 (https://feedback.fluxer.com/p/4042/system-tray-icon-not-scaled-correctly), but the latest update seems to have introduced a new scaling issue. I suspect this might be related to #4181, where the transparent margins around the monochrome icon were removed. While this may have fixed the icon being too small on GNOME, it now seems to have the opposite effect on KDE Plasma. I've attached a screenshot showing the oversized Fluxer icon next to the other system tray icons for comparison. EDIT: I know this bug report is specifically about GNOME, but I'd really like to avoid having the tray icon sizing go back and forth between GNOME and KDE Plasma. If I open a separate bug report about the icon now being too large on KDE, and the next fix makes it too small on GNOME again, we'll just end up repeating the same cycle. Ideally, the sizing should be handled in a way that works correctly on both desktop environments, rather than fixing one at the expense of the other. I'm not asking for the GNOME fix to be reverted. I'd just like to make sure that both KDE Plasma and GNOME are taken into account when adjusting the icon size. Update (November 17): After another Canary Desktop update, the monochrome tray icon is now back to the correct size on KDE Plasma. I'm not sure which change fixed it, since I couldn't find a specific tray icon scaling fix in the latest changelog. The difference is clearly visible compared to my previous screenshot. I've attached a new screenshot showing how it looks now. Hopefully, the icon sizing will remain consistent across future releases on both KDE Plasma and GNOME. :)
- For some additional context, there wasn't a burst of GIF activity before this happened. The previous GIF in that channel was about 30 minutes earlier. The first two GIFs posted afterwards failed to unfurl, while subsequent ones worked again shortly after.
- I just updated my instance and tried it again after I saw that it was added. Imo it's perfect now! I love it!! Thanks a lot!
- Yea, that would be a great addition! +1 from me for this, too! :)
- Confirmed fixed on my instance after updating to the latest release. The old
Deleted Userentries were cleaned up from/admin/users, and permanently deleting accounts now removes them without creating newDeleted Userentries. Thanks Hampus!! - The new permanent delete option is available after updating, but this still seems unresolved for accounts that were already deleted before the feature was added. Those accounts remain listed as
Deleted Userin/admin/users. Trying to permanently delete one through the Admin Panel results in409 Conflict. I also tried it through rexctl:rexctl users delete <id>which returns:DeletedUser is already being deleted.The account still remains in/admin/usersafterwards. So it looks like older deleted accounts/tombstones cannot currently be permanently removed with the new deletion flow. Ideally, these should also be removable so they no longer keep accumulating in the users overview. - Since the fix, the tray icon looks a bit too large on my KDE setup compared to the other icons. Could it be a scaling or padding issue? Edit: After the last update it looks great!
- Confirmed fixed in Web 2026.1012.154024! The splash background now covers the entire viewport while the invite card stays properly centered. Thanks for the fix! :)
- Retested this on the current Stable Web build and I can no longer reproduce the issue. I restarted the media proxy before each test to force a cold animated image transformation and repeated the test three times. The larger animated guild icon still takes around 1.56 seconds to transform:
- Test 1:
transform_ms=1555 - Test 2:
transform_ms=1554 - Test 3:
transform_ms=1560
- Test 1:
- Can reproduce H.264 selected, OpenH264 (software encoding) instead of hardware encoding. GPU: RTX 3090; Canary Desktop, Linux.
- Regarding the Calendar Channels / scheduled events part of this merged thread: One thing that would be very important for bot integrations is for calendar events to be accessible and extensible through the API. For a raid management bot I'm working on, Fluxer should ideally remain the source of truth for the event itself, while the bot only adds game-specific data. The main things we'd need are:
- stable event IDs
- API access to events and attendees/RSVPs
- gateway events when an event or RSVP changes
- stable IDs for individual occurrences of recurring events
- ideally a way for bots/apps to attach an interaction to an event
- Confirmed fixed on my self-hosted instance! System DMs now correctly display my custom instance name and avatar instead of the default Fluxer branding. Even previously sent system messages are displayed with the correct branding now. Everything looks great. Thank you so much for the fix!
- On [Self-Hosted] Custom favicon is ignored in initial HTML and `/web/favicon-32x32.png` is used insteadConfirmed fixed on my self-hosted instance!
My custom favicon now works correctly in browser tabs and bookmarks, including Firefox. The remaining Firefox issue turned out to be caused by stale favicon data in my existing Firefox profile. Rebuilding
favicons.sqliteresolved it completely. Everything is working as expected now. Thank you so much for fixing this! - One thing I think is still missing from the design plan is cross-instance guild discovery. Ideally, a user on a self-hosted instance could open Fluxer's existing Discover page and find public guilds hosted across the federated network, not just guilds hosted on their own instance. They could then join those guilds using their existing Home Instance identity. For example, if my instance only hosts two public guilds, Discover could still show eligible public guilds from fluxer.app and other federated instances. This would make self-hosted instances feel like part of one larger Fluxer network instead of isolated islands. I'm not sure whether this should use a central directory, opt-in discovery hubs, instance-to-instance discovery, or something else, but I think the user-facing experience would be very valuable. Instance and guild admins should of course be able to opt in/out of federated discovery.
- Can reproduce this issue on a self-hosted Fluxer instance running the latest available version. Environment:
- Self-hosted Fluxer
- Canary Web 2026.1008.22644
- Linux (x64), Firefox 157.0, Locale en-US
- Second user: Chrome (web client)
- Both users were using the web version, not the desktop app.
- Moving another user to a different voice channel sometimes makes them disappear from the voice channel list entirely.
- Moving them back sometimes works, but both users may then be unable to hear each other.
- Repeated channel moves can also cause the moved user to disconnect.
- Fully disconnecting and reconnecting both users temporarily resolves the issue.
Server:Connection throttled 2 Connection succeeded but attempt is stale, disconnectingLiveKit participant_joined did not match gateway state error: connection_not_found - Updated the proposal with a bit more implementation detail after looking through the current runtime branding code. The requested behaviour itself is unchanged.
- Retested after updating and I can no longer reproduce the issue. I restarted both
gifsandgifs-shardmultiple times and tested coldGetFeaturedrequests withsv-SEandpt-BR. All returned200, including cold requests taking around 1.2–1.7s. I also did a final normal test after restarting both GIF services and opening the GIF picker normally. The first request returned:Before the fix, the same cold-request scenario repeatedly resulted in 503 shard_unavailable after ~5 seconds. Looks fixed on my instance. Thanks a lot, Hampus!GET /v1/gifs/featured status: 200 durationMs: 324 - Updated the report after additional testing. The issue is not Firefox-specific. It is also reproducible in Chrome on Windows on the same self-hosted instance, and the same
window.open()workaround fixes it there as well. I updated the title, reproduction details and evidence accordingly. - Update after further investigation I was able to reproduce the issue again after a cold boot, but further testing shows that this does not appear to be a Fluxer/Flatpak or Chromium AsyncDns issue. While Fluxer was showing the gray window and repeatedly failing with
ERR_NAME_NOT_RESOLVED, DNS resolution on the host also failed:The local FRITZ!Box DNS resolver returnedresolvectl query web.fluxer.app. → Name 'web.fluxer.app.' not foundNXDOMAINfor both A and AAAA records. The FRITZ!Box is configured to use Mullvad'sextended.dns.mullvad.netresolver(194.242.2.5 / 2a07:e340::5). Direct tests against multiple resolvers showed:I repeated the test againstCloudflare 1.1.1.1 → resolves correctly Google 8.8.8.8 → resolves correctly Mullvad dns.mullvad.net → resolves correctly Mullvad adblock.dns.mullvad.net → resolves correctly Mullvad base.dns.mullvad.net → resolves correctly Mullvad extended.dns.mullvad.net → fails to resolve Mullvad family.dns.mullvad.net → resolves correctly Mullvad all.dns.mullvad.net → resolves correctlyextended.dns.mullvad.net10 times and all 10 attempts failed to resolveweb.fluxer.app. The authoritative DNS servers forfluxer.appthemselves return the correct A and AAAA records. Interestingly, Fluxer eventually recovered by itself without being restarted once DNS resolution began working again, presumably because its existing load retries eventually succeeded. So the earlier observation that disabling ChromiumAsyncDnsfixed the issue appears to have been coincidental. The current evidence points to the configured Mullvad Extended DNS resolver returning NXDOMAIN forweb.fluxer.app. I'll report this to Mullvad as well. - Additional testing: the issue currently appears to be intermittent. Initially, normal launches consistently failed with
ERR_NAME_NOT_RESOLVED, while launches withAsyncDnsdisabled and with--host-resolver-rulessucceeded immediately. However, after further testing, normal launches are now also succeeding consistently without any additional flags. So at this point I cannot confirm that disablingAsyncDnsis a reliable fix or thatAsyncDnsitself is the root cause. The original DNS resolution failures were real, but the issue is currently no longer reproducible on my system. I will update this issue if it happens again and I can reproduce it more reliably. - It looks like you're using the old Flatpak from Flathub (0.0.8). Fluxer now provides its own official Flatpak, which you can install from the download page: https://fluxer.app/download#desktop The Flatpak install command listed there is: flatpak install https://pkgs.fluxer.com/flatpak/fluxer.flatpakref Might be worth trying the current official Flatpak first and checking if the issue still occurs there.
- I really like this idea, especially for self-hosting. I think tying it to a dedicated role permission is important, otherwise users could simply mark everything as permanent and effectively bypass Media Expiry. One thing I'd especially like to see is the ability to mark an existing attachment as permanent after it has already been uploaded, e.g. via the attachment/message context menu: Keep permanently → exempts the attachment from Media Expiry Use normal expiry → puts it back under the instance's normal expiry rules A dedicated permission such as “Manage permanent media” / “Exempt attachments from expiry” would make sense to me, so this could be given to moderators without necessarily tying it to “Manage Messages”. Ideally this could also respect channel permission overrides, so for example moderators could preserve files in a resources/rules channel without having that ability everywhere. It might also be useful for these actions to appear in the audit log. “Permanent” should only mean exempt from automatic Media Expiry — deleting the actual message should still delete the attachment normally.
- Thanks! That explains the JavaScript error. However, my main concern wasn't actually the error itself, but the UX on self-hosted deployments. Since the current instance is already known from the URL, registration, email verification and login already work without ever pressing Connect. From a user's perspective, I'm not sure what purpose the editable server field and the Connect button serve in this scenario. While adding the Caddy configuration would prevent the JavaScript error, it would still leave the same UX question open. I can definitely see this UI making sense for a future generic/official Fluxer client that can connect to arbitrary Fluxer instances, but I'm not sure it makes sense for dedicated self-hosted deployments. Since this issue seems to be a duplicate regarding the JavaScript error, would it make more sense to open a GitHub Discussion (or a separate feature request) to discuss the UX aspect?
- This would be especially useful if invite-based registrations could be configured with:
- maximum number of uses
- expiration dates
- optional role assignment
- optional email requirement
- I believe I'm seeing the same issue on a self-hosted instance. While the issue was active I inspected the DOM and React tree and found that the member list was rendering skeleton entries rather than actual member entries. The sidebar showed correct group names and counts (e.g. "Administrator — 1", "Offline — 1"), but member rows remained stuck as skeletons. Switching to another channel or community immediately caused the member list to render correctly again. I also observed one case where the member list briefly rendered correctly and then returned to the skeleton state less than a minute later without any further interaction. Environment:
- Firefox on Linux
- Self-hosted instance