~dmj726/novacut/audiorender : changes

~dmj726/novacut/audiorender » Changes from revision 411

From Revision 411 to 392

Rev	Summary	Authors	Tags	Date
411	audiorender.py: edit final buf in slice and correct off-by-o... audiorender.py: edit final buf in slice and correct off-by-one error to produce right number of samples	David Jordan		7 years ago
410	audiorender.py: produce correct audio timestamps and output ... audiorender.py: produce correct audio timestamps and output (almost) correct number of samples	David Jordan		7 years ago
409	initial WIP audio-only renderer initial WIP audio-only renderer	David Jordan		7 years ago
408	Bump version to 16.05.0~alpha Bump version to 16.05.0~alpha	Jason Gerard DeRose		7 years ago
407	Release 16.03.0 Release 16.03.0	Jason Gerard DeRose	16.03.0	8 years ago
406	debian/control: Build-Depend on pyflakes3 \| pyflakes; requir... debian/control: Build-Depend on pyflakes3 \| pyflakes; require dmedia >= 16.03	Jason Gerard DeRose		8 years ago
405	Bump version to 16.03.0~alpha Bump version to 16.03.0~alpha	Jason Gerard DeRose		8 years ago
404	Drop V0 to V1 migration functionality. Drop V0 to V1 migration functionality. This change drops the `novacut-v0-v1-upgrade` script along with the `novacut.migration` and `novacut.tests.test_migration` modules. The V1 Dmedia hashing protocol, V1 Dmedia document schema, and V1 Novacut document schema were finalized over 29 months ago (on June 1st 2013): https://launchpad.net/filestore/+announcement/11554 This has given folks ample time to migrate any existing V0 libraries to V1. However, there is a burden in continuing to support this migration path, especially in the `filestore` component. Because V1 switched from RFC-3548 Base32 encoding to our Dbase32 encoding, the V0 and V1 Dmedia hashing protocols could not be supported simultaneously in the same file-store layout. The V0 to V1 migration had extra complexity that we don't expect in any future migrations. So it's a good time to drop this baggage from `filestore`, which requires likewise dropping the V0 to V1 schema migration functionality from Dmedia and Novacut. Note the plan is to indefinitely support the V1 hashing protocol, and to indefinitely support migration from the V1 document schema to whatever document schema version Dmedia and Novacut currently use.	Jason Gerard DeRose		8 years ago
403	Doodle on new generalized frame-server using Degu Doodle on new generalized frame-server using Degu	Jason Gerard DeRose		8 years ago
402	Drop id, src from render.Slice namedtuple as they are from t... Drop id, src from render.Slice namedtuple as they are from the renderservice domain	Jason Gerard DeRose		8 years ago
401	Bump version to 15.11.0~alpha Bump version to 15.11.0~alpha	Jason Gerard DeRose		8 years ago
400	Bump version to 15.10.0~alpha Bump version to 15.10.0~alpha	Jason Gerard DeRose	15.10.0	8 years ago
399	Sundry cleanup, add some missing unit tests Sundry cleanup, add some missing unit tests	Jason Gerard DeRose		8 years ago
398	Drop support for Gst 1.2, 1.4; drop Gst 1.2 hacks Drop support for Gst 1.2, 1.4; drop Gst 1.2 hacks	Jason Gerard DeRose		8 years ago
397	Doodle on some deep validation helpers, deep-test.py script Doodle on some deep validation helpers, deep-test.py script	Jason Gerard DeRose		8 years ago
396	re-render-test.py should call Renderer.destroy(), Validator.... re-render-test.py should call Renderer.destroy(), Validator.destroy()	Jason Gerard DeRose		8 years ago
395	ui/bucket.js: drop stop=(frames - 2) Gst 1.2 hack on new sli... ui/bucket.js: drop stop=(frames - 2) Gst 1.2 hack on new slices, let's not even go there	Jason Gerard DeRose		8 years ago
394	Replace EOS handling mess with something good. Replace EOS handling mess with something good. Previous attempt was to handle EOS events and internal "we want to stop here" conditions (aka, work-arounds for Gst 1.2) using the same unified code paths for Gst 1.2 through 1.6. Unfortunately, this turned out to be an unholy racy mess that introduced regressions from Gst 1.2 to shining Gst 1.6. Fortunately, with Gst 1.4 and newer, EOS doesn't fire early, works with rather perfect reliability. We can count on it to fire at the true end of a non-simple seek, not before; so there's no reason to 2nd guess it with a GLib.idle_add(self.complete, True) from within the appsink "new-sample" event handler once we reach the final frame in a slice. So... this new attempt was first written to be a simple, correct solution for Gst >= 1.4, and seems to have solidly fixed the regressions there. Then as needed, horrible special case hacks were grafted on for Gst 1.2, but these code paths are only taken when USE_HACKS is True. There may be some remaining regressions with Gst 1.2, but we're mostly back to where we were. As very soon we'll be dropping support for Trusty for new Novacut development, will only be targeting Wily and newer (GStreamer 1.6 and newer), this is the direction we want to be heading, even if there is still some Gst 1.2 baggage in the mean time.	Jason Gerard DeRose		8 years ago
393	Serval tweaks lower renderer memory usage, biggest win being... Serval tweaks lower renderer memory usage, biggest win being dropping x264enc rc-lookahead from 40 to 20	Jason Gerard DeRose		8 years ago
392	pyflake fixes in test_thumbnail.py pyflake fixes in test_thumbnail.py	Jason Gerard DeRose		8 years ago

Older »