Skip to content

Releases: membraneframework/membrane_core

v1.1.1

24 Jul 15:45
3caf044
Compare
Choose a tag to compare

Bugfixes

  • Fix 'table identifier does not refer to an existing ETS table' error when inserting metrics into the observability ETS by @mat-hek in #835
  • DistributedAtomic.Worker implements sub_get as a call by @philipgiuliani in #837

Other pull requests

New Contributors

Full Changelog: v1.1.0...v1.1.1

v1.1.0

27 Jun 12:47
5f53b95
Compare
Choose a tag to compare

Biggest changes

  • Introduction of new callbacks in Bins and Pipelines: handle_child_setup_completed/3 and handle_child_playing/3
  • Improvements of :auto flow control mechanism, including queueing items, when processing them is not demanded in the specific moment

All changes

New Contributors

Full Changelog: v1.0.1...v1.1.0

v1.1.0-rc1

05 Jun 10:36
14d8018
Compare
Choose a tag to compare

What's Changed

New Contributors

Full Changelog: v1.1.0-rc0...v1.1.0-rc1

v1.1.0-rc0

08 Apr 12:10
fb10564
Compare
Choose a tag to compare

This release contains:

  • Upgrade in flow control mechanism - queuing buffers, when demand is low in flow_control: :auto.
  • Few bug fixes.
  • Deprecation of handle_spec_started/3 callback and Membrane.Testing.Pipeline.message_child/3 in favor of Membrane.Testing.Pipeline.notify_child/3.
  • Bump dependency to Ratio to ~> 3.0 or ~> 4.0.
  • Docs upgrade.

All changes

New Contributors

Full Changelog: v1.0.1...v1.1.0-rc0

v1.0.1

22 Feb 14:59
2eacd1a
Compare
Choose a tag to compare

This release contains some bug fixes and small improvements. The most important are:

  • Fix for a bug in the timer that made it tick two times too slow sometimes (non-deterministically). Fixed in #685.
  • There's a new field crash_reason in Pipeline's and Bin's callback context that can be used in the handle_crash_group_down callback, added in #720.
  • Unrecognized messages now trigger a warning if a component doesn't have the handle_info callback implemented, just like in GenServer. See #680.

All changes

New Contributors

Full Changelog: v1.0.0...v1.0.1

v1.0.0

16 Oct 14:52
506a41e
Compare
Choose a tag to compare

Finally, we have released v1.0.0 of Membrane Core! 🎉🎉🎉 Comparing to v0.12.9, this releases contains many changes in API and framework private modules. We wrote a migration guide helping you adjust your code to Membrane Core v1.0.0. Let's take a look, at what changes are introduced with this release:

Improvements and optimizations in the demands mechanism

Release v1.0.0 contains a lot of changes in Membrane private modules responsible for managing the backpressure mechanism. The two biggest changes are:

  • Implementation of :push effective flow control in elements containing pads with flow_control: :auto - it allows avoiding work related to the calculation of the value of auto-demand in input pads with flow_contro: :auto, when a pipeline topology allows us to skip it without damage of backpressure mechanism. This change also fixes the problem with toilet overflow error occurring far from the element, which works too slowly.
  • Passing the value of demand using :atomics, instead of message passing - this change is a big performance and memory optimisation and allowed us to make auto-demands more precise in the amount of demanded data.

Support for kino_membrane

This release brings compatibility with kino_membrane, the tool for visualization and introspection of Membrane pipelines in Livebook.

Sending end of stream, even if there was no start of stream before

Since now, we can send end of stream on pads, even if we haven't sent start of stream there yet. Moreover, before unlinking an input pad, every element will always execute handle_end_of_stream for this pad, no matter, if it received start of stream earlier, or not. This will happen before calling handle_pad_removed/3 in the case of dynamic pads and on entering terminating state, in the case of static pads.

Removing handle_process_list/4 and handle_write_list/4 callbacks

Since v1.0.0, you have to handle every single buffer separately in handle_buffer/4 callback, instead of handling whole list of buffers in this in handle_process_list/4 or handle_write_list/4.

API improvements

  • handle_process and handle_write are replaced with handle_buffer for better consistency
  • :mode and :demand_mode options in pad definition are replaced with :flow_control, which may be :push, :auto or :manual and defaults to :auto whenever available.
  • Callback contexts are now maps instead of structs and they're documented in a single module per component, so it's more straightforward to figure out what they contain

All changes

Breaking changes

Non-breaking changes

Bug fixes

Full Changelog: v0.12.9...v1.0.0

v1.0.0-rc1

02 Oct 12:41
fd16779
Compare
Choose a tag to compare

Here comes the second release candidate for Membrane Core 1.0.0! Compared to v1.0.0-rc0, this release does not contain a lot of changes in API, but it introduces many bug fixes, improvements and optimizations in the Membrane private modules. We wrote a migration guide, helping you in adjusting your code to this version of membrane_core. But now, let's take a look at the summary, of what changes we have just introduced:

Improvements and optimizations in the demands mechanism

Release v1.0.0-rc1 contains a lot of changes in Membrane private modules responsible for managing the backpressure mechanism. The two biggest changes are:

  • Implementation of :push effective flow control in elements containing pads with flow_control: :auto - it allows avoiding work related to the calculation of the value of auto-demand in input pads with flow_contro: :auto, when a pipeline topology allows us to skip it without damage of backpressure mechanism. This change also fixes the problem with toilet overflow error occurring far from the element, which works too slowly.
  • Passing the value of demand using :atomics, instead of message passing - this change is a big performance and memory optimisation and allowed us to make auto-demands more precise in the amount of demanded data.

Support for kino_membrane

This release brings compatibility with kino_membrane, the tool for visualization and introspection of Membrane pipelines in Livebook.

Sending end of stream, even if there was no start of stream before

Since now, we can send end of stream on pads, even if we haven't sent start of stream there yet. Moreover, before unlinking an input pad, every element will always execute handle_end_of_stream for this pad, no matter, if it received start of stream earlier, or not. This will happen before calling handle_pad_removed/3 in the case of dynamic pads and on entering terminating state, in the case of static pads.

Removing c:Membrane.Element.WithInputPads.handle_buffers_batch/4

This release reverts support for this callback, so if you have used it, you have to rewrite your element to use c:Membrane.Element.WithInputPads.handle_buffer/4 instead.

Crash groups in bins

Until now, it was possible, to create a crash group only from the level of pipelines. v1.0.0-rc1 introduces support for doing it also from bins!

All changes

Breaking changes

Non-breaking changes

Bug fixes

New Contributors

Full Changelog: v1.0.0-rc0...v1.0.0-rc1

v0.12.9

14 Sep 15:01
Compare
Choose a tag to compare

What's Changed

  • Add :pause_auto_demand and :resume_auto_demand actions. #586
  • Fix process leak in starting clocks. #594
  • Add child exit reason to the supervisor exit reason. #595

Full Changelog: v0.12.8...v0.12.9

v0.12.8

24 Aug 12:51
Compare
Choose a tag to compare

What's Changed

  • Move metrics ets creation to stalker by @mat-hek in #589
  • Refactor code related to crash groups and linking mechanism by @FelonEkonom in #574
  • Fix specs for default pipeline functions by @FelonEkonom in #585
  • Fix stream format error message, fix race condition in handling child pad removed by @FelonEkonom in #591

Full Changelog: v0.12.7...v0.12.8

v0.12.7

21 Jul 11:29
Compare
Choose a tag to compare

What's Changed

Full Changelog: v0.12.6...v0.12.7