1190472 - MediaRawDataQueue::Push/PushFront should take && arguments to reduce refcounting

Status: RESOLVED FIXED

(Core :: Audio/Video, enhancement, P5)

Description

[Aryeh Gregor (:ayg) (no longer with Mozilla)]

(Nathan Froyd [:froydnj][:nfroyd] from bug 1179451 comment #32)
> I wonder if these PushFront methods should take && arguments as well or
> instead of pointer arguments, just to avoid unnecessary reference-counting. 
> File a followup bug?

Comment 1

[Nathan Froyd [:froydnj]]

I have patches for this, I'll post them tomorrow.

Comment 2

[Nathan Froyd [:froydnj]]

Attachment: part 1 - improve MediaRawDataQueue's reference-counting behavior

OK, so "tomorrow" turned into "two weeks", but this is not exactly
high-priority stuff.

Sometimes when we push onto the queue, we don't need to hold a reference
afterwards.  In that case, we can pass the reference in and avoid
unnecessary reference counting.

Comment 3

[Nathan Froyd [:froydnj]]

Attachment: part 2 - delete unused MediaRawDataQueue::Push method

We don't call it, and it's not particularly efficient anyway.

Comment 4

[Nathan Froyd [:froydnj]]

Attachment: part 3 - optimize pushing an entire queue onto MediaRawDataQueue

In the one place we push a MediaRawDataQueue onto a MediaRawDataQueue,
we don't use the pushee aftewards.  It's more efficient to indicate that
by using Move(), and we can then save on reference-counting things
needlessly.

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/e0d5a28ba9d4
https://hg.mozilla.org/integration/mozilla-inbound/rev/3c4316e8d12a
https://hg.mozilla.org/integration/mozilla-inbound/rev/0e80a5538a67

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/e0d5a28ba9d4
https://hg.mozilla.org/mozilla-central/rev/3c4316e8d12a
https://hg.mozilla.org/mozilla-central/rev/0e80a5538a67

Comment 7

[Jean-Yves Avenard [:jya]]

Part3 incorrectly pushed to the wrong end of the queue. this broke webm playback whenever this code was called.

Comment 8

[Nathan Froyd [:froydnj]]

(In reply to Jean-Yves Avenard [:jya] from comment #7)
> Part3 incorrectly pushed to the wrong end of the queue. this broke webm
> playback whenever this code was called.

Doh.  Sorry about that. :(