Now you can remove the initial setting of m_device as well.

In D10236#199252, @fvogt wrote:

Now you can remove the initial setting of m_device as well.

I thought about it while doing the change. I decided against it for the following reasons:

• it would need to be inited with null instead
• the norm is not having this problem, the switch event is an exception
• device would have to lose the const-ness, the const_cast hack is in my opinion only a valid solution for exceptions, for the standard case it would have to be changed.

In D10236#199259, @graesslin wrote:

In D10236#199252, @fvogt wrote:

Now you can remove the initial setting of m_device as well.

I thought about it while doing the change. I decided against it for the following reasons:

• it would need to be inited with null instead

Alternatively just call device() once explicitly in the constructor to avoid code duplication.
That way the lazyness of device() is lost, but that's not a goal anyway.

• the norm is not having this problem, the switch event is an exception

It's not, this can happen to literally any event if they get queued together.

• device would have to lose the const-ness, the const_cast hack is in my opinion only a valid solution for exceptions, for the standard case it would have to be changed.

Just set m_device as mutable instead, that's what it's intended for.

I'm actually in favor of a tar respin for 5.12.0 in this case. What do you think?

In D10236#200072, @fvogt wrote:

I'm actually in favor of a tar respin for 5.12.0 in this case. What do you think?

I would not have asked for it, but I agree. @bshah can you please take care of it?

In D10236#200079, @graesslin wrote:

I would not have asked for it, but I agree. @bshah can you please take care of it?

Yeah give me half an hour

Ergh, can you commit it though?