- Kaunas, Lithuania
- https://aras-p.info/
- Joined on
2022-01-20 -
-
The current PR state as-is right now still does not fix the "some sound strips get negative length" issue. Did you check with my file from "much smaller repro for this problem, attached" comment? The active strip there is still "broken" with this PR.
What is the reason why this code path proceeds to setup retiming keys that effectively do "no retiming", instead of returning, without setting up retiming stuff? If there is one, it would be good to state the reason in a comment.
Should version_seq_fix_broken_sound_strips code from versioning_300.cc be also moved or removed? Right now it is under 306, 9 version.
@David-Garcia-1 not yet! if you want to take a go at it, be my guest
Explanation makes sense to me. I know nothing about that code though :)
I was able to make a much smaller repro for this problem, attached.
Here's how it looks in 3.3 - 3.6 (top strip length 94):
An…
Could be the reason for this too: blender/blender#118719
Very unlikely. But on that issue, no one else could reproduce it, and you never attached files…
I haven't looked at the code yet (but started builds + tests), but what would immediately help is: if you edited the PR description to describe what is being done, what bugs/problems is it…
Yes, Windows overall does not allow drag-and-drop between applications of different privilege levels. Not much (or rather, nothing) that Blender can do here.
Should this have similar logic as do_versions_sequencer_init_retiming_tool_data in versioning_300.cc, where it also checks for SEQ_retiming_is_allowed, and after adjusting the retiming stuff it sets speed_factor to 1.0?
Hmm does not seem to help much. It changes the affected strips length, but from the wrong negative number to a different wrong negative number.
I have tried making a smaller repro than whole…
@iss ah indeed, looks like the Sprite Fright Edit file gets a bunch of the strips "wrong" somewhere between Blender 3.6 (correct) and 4.0 (wrong). Looks like it is sound strips that used to have…
And the file is broken, because of bad versioning, so that should be fixed.
Ah, you mean there is missing versioning somewhere in the code? Can you point me to where it might be? I could try…
Well, default developer config has WITH_ASSERT_ABORT, so that assert firing quits blender. And Sprite Fright Edit is the primary file that I'm using for VSE testing, so yes this is annoying at…
