Fix livestreams in pushback mode incorrectly writing to queue archive. #180

Closed
opened 2025-09-17 07:59:27 -04:00 by rainbownapkin · 9 comments
rainbownapkin commented 2025-09-17 07:59:27 -04:00 (Migrated from gitlab.com)

Livestream marker on the schedule displays incorrectly for previous days during a multi-day stream.

Also streaming in pusback mode into the next day completely fucks the schedule, causing it to reschedule everything immediately.

Livestream marker on the schedule displays incorrectly for previous days during a multi-day stream. Also streaming in pusback mode into the next day completely fucks the schedule, causing it to reschedule everything immediately.
rainbownapkin commented 2025-09-17 07:59:27 -04:00 (Migrated from gitlab.com)

added #164 as parent issue

added #164 as parent issue
rainbownapkin commented 2025-09-17 08:00:21 -04:00 (Migrated from gitlab.com)

changed the description

changed the description
rainbownapkin added a new dependency 2025-09-25 05:39:24 -04:00

results from a quick test:

started s4 at 1030

started stream during s04e04 @ 11:55PM (sim)

ended stream during s04e06 @ 12:25AM (sim)

results:

Mostly normal?

Original S04 item expanded within queue window to take up entire schedule but start/end times looked right?

results from a quick test: started s4 at 1030 started stream during s04e04 @ 11:55PM (sim) ended stream during s04e06 @ 12:25AM (sim) results: Mostly normal? Original S04 item expanded within queue window to take up entire schedule but start/end times looked right?

started s2 at 1040 - Fri Oct 24 07:19:04 AM EDT 202

started stream during s02e04 @ 11:55PM (sim)

ended stream during s02e05 @ 12:15AM (sim)

results:

Mostly normal? Media seems to resume and pushback normally, at least the part that runs after the livestream....

The orignal instance of the ep which was played over (s02e04) was, however, improperly written to the schedule archive.

Notes from dumping client.queue in the browser:

  • Items from before the stream seem flipped in order afterwords. (We need to test if this happens every time, or just for over-night streams)
  • Original instance which is miss-rendered in the queue, seems to look relatively normal in the client.queue outisde of being out of order. Maybe this is an issue with rendering media out of order? (not that we should ever enter that condition)
started s2 at 1040 - Fri Oct 24 07:19:04 AM EDT 202 started stream during s02e04 @ 11:55PM (sim) ended stream during s02e05 @ 12:15AM (sim) results: Mostly normal? Media seems to resume and pushback normally, at least the part that runs after the livestream.... The orignal instance of the ep which was played over (s02e04) was, however, improperly written to the schedule archive. Notes from dumping client.queue in the browser: - Items from before the stream seem flipped in order afterwords. (We need to test if this happens every time, or just for over-night streams) - Original instance which is miss-rendered in the queue, seems to look *relatively* normal in the client.queue outisde of being out of order. Maybe this is an issue with rendering media out of order? (not that we should ever enter that condition)

started at gb trailer (7:09) - Everythings but last played is in order

ended at gc intro (07:16) - everything old is out of order, randomly requed everything but ghostbusters after ghostbusters instead of continuing proper.

It seems like this is NOT related to an issue with overnight pushback. Pushback mode is just busted lol

started at gb trailer (7:09) - Everythings but last played is in order ended at gc intro (07:16) - everything old is out of order, randomly requed everything but ghostbusters after ghostbusters instead of continuing proper. It seems like this is NOT related to an issue with overnight pushback. Pushback mode is just busted lol

^Repeated results in similar test, this is a VERY similar bug to the "overnight livestream bug" we saw during the v0.3-indev public crash test, and absolutely repeatable, and happens regardless of the time.

This, paired with the weird client-side rendering bugs related to over-night livestreams, may have caused me to infer correlations that don't exist. ONE OF THESE THINGS IS NOT LIKE THE OTHER!!! Two bugs can happen at once! this part of the bug happens regardless of time, which makes sense. The server doesn't give a flying fuck what day it is, it works in epochs.

TARGET THIS!!!!

^Repeated results in similar test, this is a *VERY* similar bug to the "overnight livestream bug" we saw during the v0.3-indev public crash test, and absolutely repeatable, and happens regardless of the time. This, paired with the weird client-side rendering bugs related to over-night livestreams, may have caused me to infer correlations that don't exist. ONE OF THESE THINGS IS NOT LIKE THE OTHER!!! Two bugs can happen at once! this part of the bug happens regardless of time, which makes sense. The server doesn't give a flying fuck what day it is, it works in epochs. TARGET THIS!!!!
rainbownapkin changed title from Fix overnight livestream related bugs. to Fix livestreams in pushback mode incorrectly writing to queue archive. 2025-10-24 19:32:05 -04:00

Overwrite preforms much better in this test, but still puts the start video and live stream out of order, and before everything else in the queue.

We should dump queue state from the server to confirm these findings, before continuing.

Overwrite preforms much better in this test, but still puts the start video and live stream out of order, and before everything else in the queue. We should dump queue state from the server to confirm these findings, before continuing.

The issue with pushback re-queueing old items has been solved by: 37990ff8c3, as the RAM-backed queue was holding on to stale items which had already been archived to the DB.

This caused the logic to re-add them when it iterated the RAM-backed queue upon the streams end.

Overwrite seems to preform just fine too.

Looking at the server-side queue from the dump debug command, everything seems to be fine on. Though the client side reports out of order, it doesn't seem to be a big issue.

Looking into queue.prepQueue to see whats up with the server sending the client out-of-order queue information when the server-side is all set.

The issue with pushback re-queueing old items has been solved by: 37990ff8c3, as the RAM-backed queue was holding on to stale items which had already been archived to the DB. This caused the logic to re-add them when it iterated the RAM-backed queue upon the streams end. Overwrite seems to preform just fine too. Looking at the server-side queue from the dump debug command, everything seems to be fine on. Though the client side reports out of order, it doesn't seem to be a big issue. Looking into queue.prepQueue to see whats up with the server sending the client out-of-order queue information when the server-side is all set.

Fixed out of order queues being sent off to clients: dd66601f0d

Fixed out of order queues being sent off to clients: dd66601f0d
Sign in to join this conversation.
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Blocks
#164 Tweaks and Fixes
rainbownapkin/canopy
Reference: rainbownapkin/canopy#180
No description provided.