Store dedicated transform listener thread as a std::unique_ptr

[allenh1]

Signed-off-by: Hunter L. Allen hunterlallen@protonmail.com

[tfoote]

Follow up to #108 (comment)

• Linux
• Linux-aarch64
• macOS
• Windows

[Karsten1987]

I don't see the big benefit of converting the raw pointer to a unique_ptr if not putting the rest of the destructor logic in a custom deleter and thus having the unique ptr taking care of the dedicated thread itself.

[Karsten1987]

@allenh1 is there any update from your side? Would be great to have this PR landed before dashing.

[allenh1]

@Karsten1987 my bad, I'll get to this ASAP.

[allenh1]

@Karsten1987 ptal -- this is kind of clean, but not my favorite patch. I couldn't get the std::unique_ptr to cooperate without declaring it with two template args (since it will default initialize it in other ways). I ended up with a using clause; let me know if you want me to kill it and, if so, I'll just inline the definitions.

[Karsten1987]

Can we delete using_dedicated_thread_ then? It looks pretty unnecessary to me. With it, we can also delete the private method dedicatedListenerThread which doesn't do anything.
This should IMO not break API but max ABI, which at this point should still be ok for dashing.

@tfoote opinions?

[Karsten1987]

my personal style would just be to put the using just above the first (and only) declaration of this type in line 115.

[tfoote]

Since it's only used once that is a lot easier to read.

[tfoote]

Yes, we can clear the old flag using_dedicated_thread_ and the old unused function dedicatedListenerThread that's replaced by the lambda. We should move the TODO from that method up to the lambda to remind us to come back and implement it with callback groups when callbackgroups become available.

[tfoote]

Since it's only used once that is a lot easier to read.

[allenh1]

@Karsten1987 / @tfoote Could you please spin up the CI for me?

[tfoote]

lgtm, one last ci run:

• Linux
• Linux-aarch64
• macOS
• Windows

[allenh1]

Woot