lp:~icecast/icecast/kh

Created by Jeff Waugh on 2014-09-13 and last modified on 2019-10-19
Get this branch:
bzr branch lp:~icecast/icecast/kh

Branch merges

Related bugs

Related blueprints

Branch information

Owner:
Icecast
Project:
Icecast
Status:
Development

Import details

Import Status: Reviewed

This branch is an import of the HEAD branch of the Git repository at https://github.com/karlheyes/icecast-kh.git.

The next import is scheduled to run in 6 minutes.

Last successful import was 5 hours ago.

Import started 5 hours ago on izar and finished 5 hours ago taking 20 seconds — see the log
Import started 11 hours ago on izar and finished 11 hours ago taking 20 seconds — see the log
Import started 17 hours ago on alnitak and finished 17 hours ago taking 25 seconds — see the log
Import started 23 hours ago on izar and finished 23 hours ago taking 20 seconds — see the log
Import started on 2019-12-15 on alnitak and finished on 2019-12-15 taking 25 seconds — see the log
Import started on 2019-12-14 on izar and finished on 2019-12-14 taking 20 seconds — see the log
Import started on 2019-12-14 on alnitak and finished on 2019-12-14 taking 20 seconds — see the log
Import started on 2019-12-14 on izar and finished on 2019-12-14 taking 20 seconds — see the log
Import started on 2019-12-14 on izar and finished on 2019-12-14 taking 20 seconds — see the log
Import started on 2019-12-13 on alnitak and finished on 2019-12-13 taking 20 seconds — see the log

Recent revisions

871. By Karl Heyes <email address hidden> on 2019-10-19

missing on-demand setting in master block.

This is just like an alias for the relays-on-demand setting. If the multiple master
facility gets done then it can be used specifically for that master.

870. By Karl Heyes <email address hidden> on 2019-10-18

fix race case, adjust locking area when listeners move or drop off.

In some cases, when moving to another worker or leaving a stream, there is a point where
a source is referred to but not locked as it should be. By changing the locking constraints
a lock/data is being maintained even if the source is to be freed up.

869. By Karl Heyes <email address hidden> on 2019-09-22

add some tolerance when pruning queue

normally it should not matter but there is a race case between source client pruning and
a listener detaching from queue when the lock is re-acquired for updating.

868. By Karl Heyes <email address hidden> on 2019-09-10

prevent incorrectly defined mime types from breaking relays

This is not usually a problem, but if one is define and it is not a recognised type then some
initialisation of relays can be missed out on retrying. Initially when relays a created they
are setup and work fine, but on retry, the structures and settings are not cleared out as type
checking for listeners is maintained but initialisation needs to be done for certain things
like metadata inserts.

This mismatch was tripping a bug here causing metadata alignment problems. Now we verify the
specified type before overriding.

867. By Karl Heyes <email address hidden> on 2019-09-09

enforce a / on local-mount for relays.

if that is missing then it can lead to a mismatch between relay and source and mountpoint
names. possible crash or hang on reload.

866. By Karl Heyes <email address hidden> on 2019-09-09

fix silly return value, largely ignored for now

865. By Karl Heyes <email address hidden> on 2019-09-08

avoid generic reader reschedule if block has been added. and clean up silly log message

silly log message was from the intro file load when there isn't one to load

the skip setting was adversely affecting the reader part of certain high bandwidth streams. These are
typically Mb/s. The rescheduling ended up increasing in certain cases to a point where lowering it
would take it so long that the source would drop. This has shown when the source does not send anything
straight away then starts blasting the content

864. By Karl Heyes <email address hidden> on 2019-09-07

enforce a range check in the queue sizing.

seems like the assessed range could get messed up because clients could drop out and
any time and leave the targeting too short. You could also get a lagging client
also stretching the queue but that is less of an issue as that tends to resolve on a
later interation. The former though could cause excessive purging.

863. By Karl Heyes <email address hidden> on 2019-09-06

closed socket and block sizing update changes

the read test for closed proxy socket was far too frequent on high bitrates, so increase the
frequency of them and make sure the nothing to do case also follows that same routine.

make the automatic block size increase at a lower threshold, this should increase the client
reschedule time slightly. Should only apply for high bitrates.

862. By Karl Heyes <email address hidden> on 2019-05-29

the internal incoming rate setting not properly set with relays

the stats are ok as those are independently assessed but there are certain checks based
on the internal setting that reference this and was only being done for source clients.

Branch metadata

Branch format:
Branch format 7
Repository format:
Bazaar repository format 2a (needs bzr 1.16 or later)
This branch contains Public information 
Everyone can see this information.

Subscribers