Sadly this didn't make it into Blender 3.0.
It will be in the 3.1 release.

LGTM.
I guess you could include this in the 2.93 LTS release as well?

So you are saying that this fix doesn't actually fix the problem?

I can reproduce this issue.
It works with Eevee, not cycles.

LGTM :)

This is a known issue with depsgraph and the rigidbody simulation system.
Undoing in certain cases will trigger a crash.

Other than my nitpicks, LGTM!

Updated with windows related changes.

Also note that I will merge the fix into the 2.93 branch, so this should be in the next 2.93 release.
If that wasn't clear. Thanks again for pointing this out.

The general idea is that you file out one report for all blender versions that are affected.

Thanks for pointing out that I forgot to merge this fix into the 2.93 branch.

Adding Ray as well for the windows part that I haven't changed yet.
Is the install for windows needed here?

If someone else hasn't already reported this issue, then you should report it as an other issue, yes.

I'm saying that the invalid geometry generated by the boolean modifier is an other issue.
You reported that blender crashed when quadriflow was used on bad meshes with zero length edges.

In either case, the issue with generating invalid meshes is an other issue. So the reported crash has been "fixed", so I'll close this.

Did you try the latest 3.0 version?

The latest file you attached works in master as well (because it doesn't have zero length edges).

The model has several zero length edges which quadriflow can't handle.
(Use "merge by distance" to remove these and it will not crash anymore)

This approach looks good to me.

@Clément Foucault (fclem) Could you take a look?

In D12729#334201, @Peter Fog (tintwotin) wrote:

If it is out of the question of getting the multi tool to work on the selection, I don't know if it is allowed to force change a tool change(in the tool bar), but in a case like the video, when selecting more strips then the multi tool could be disabled and the active tool could be changed to the selection tool?

We don't do automatic tool switching in these cases.
It would be very annoying having to reselect the bounding box tool is you by mistake select multiple strips.

Updated patch with the latest feedback.

Forgot to add Campbell.
@Campbell Barton (campbellbarton) does the 2d gizmo drawing changes look good?
I removed the modal calculations in the drawing code as it seems to me that the result there was always overwritten by the "pre draw" callback functions either way.

I've talked to Sybren and Brecht.

Yes, I tested this with upstream libs on my linux machine as well. So I think it is on our end too.

LGTM

Besides my comments, I want to rename "overshoot" as well.
Basically because I think overshoot is not describing what it happening.

To me this sounds like failing hardware. As per the fedora FAQ about these types of errors: link

We need to do a deep clean/refactor of our rigidbody code to properly fix some other crashing issues that are there because of some very bad design choices in the current code that can't be fixed.
So as Sergey pointed out, I don't either think this will be a high priority issue because of the amount of planning and time this need to be fixed properly.

@Sergey Sharybin (sergey) Do you have any objections to this?
I don't think that a small memory leak when opening the file in older versions is anything to worry about.

I think it looks good to me except a few things that I created a patch for:

I'm on vacation for two weeks, so I'll review and test this when I get back.

In D12325#320236, @Richard Antalik (ISS) wrote:

Only now I noticed, that this will produce error in older blender versions error: negative users in strip and therefore this will cause some small-ish memory leak. I am not quite sure whether this would be a cause for concern though.

In T90676#1211253, @Brecht Van Lommel (brecht) wrote:

I don't see how SDL solves any of the specific issues we are having now? Did you see any mechanism where it helps us avoid linking the Blender executable to X11 or Wayland libraries and decide to load them at runtime only?

Right, I also think that those two points will require a lot of work.

In rBc634d859b216602d3d479966a5e48a433777ddce#313422, @Peter Fog (tintwotin) wrote:

The light center of the waveform makes the transform numbers hard to read:

I think we should probably add text outlines to solve this.

A few thoughts:

In D12034#319113, @Christian Rauch (christian.rauch) wrote:

I created a design task at https://developer.blender.org/T90676. There seem to be no objections against the proposal, but also no explicit confirmation of the approach.

Do you think this is sufficient? Shall we now go on with the proposed change?

In D11917#318533, @Joerg Mueller (nexyon) wrote:

I know what that is and this is on my list to be fixed, so don't bother investigating it.

Nice! Thanks for the heads up :)

In D11917#317970, @Joerg Mueller (nexyon) wrote:

I just committed two fixes for pulseaudio, I'm not sure if they help, please check out a recent build of master (3.0.0 alpha), e.g.: https://builder.blender.org/download/experimental/blender-3.0.0-alpha+master.a217e043be2d-linux.x86_64-release.tar.xz

Ah, thanks! :)

I'll just tag @Bastien Montagne (mont29) as he wanted to do a double check on the for_each logic