1201277 - Use a lambda instead of a local class as the condition for APZCTreeManager's calls to BreadthFirstSearch()

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Botond Ballo [:botond]]

APZCTreeManager makes two calls to BreadthFirstSearch():

  https://dxr.mozilla.org/mozilla-central/rev/fb720c90eb49590ba55bf52a8a4826ffff9f528b/gfx/layers/apz/src/APZCTreeManager.cpp#1560

  https://dxr.mozilla.org/mozilla-central/rev/fb720c90eb49590ba55bf52a8a4826ffff9f528b/gfx/layers/apz/src/APZCTreeManager.cpp#1579

Both calls take a 'condition' argument, which tells the search when to stop (when a node that meets the condition is found).

We currently use local classes for both conditions. Using a C++11 lambda would be nicer, but we couldn't do so at the time we wrote this code due to a bug in our static analysis plugin (bug 1170388) that caused a false-positive static analysis error for a use of a lambda there.

Now that this static analysis bug is fixed, we should update the calls to use C++11 lambdas.

This is a good first bug for someone new to the Mozilla codebase (less so for someone new to C++).

Comment 1

[Michael Pruett]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=d41056e89c89

Comment 2

[Michael Pruett]

Attachment: Patch

I've updated APZCTreeManager replacing functors with lambdas where appropriate.

Comment 3

[Botond Ballo [:botond]]

Comment on attachment 8656991 [details] [diff] [review]
Patch

Review of attachment 8656991 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks, Michael! It looks good.

I'm going to do a static analysis build on Try before landing, just in case:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=42d1d6cd7083

Comment 4

[Botond Ballo [:botond]]

(In reply to Botond Ballo [:botond] from comment #3)
> I'm going to do a static analysis build on Try before landing, just in case:
> 
> https://treeherder.mozilla.org/#/jobs?repo=try&revision=42d1d6cd7083

Looks good, landed:

https://hg.mozilla.org/integration/mozilla-inbound/rev/0fe11ce43945

Comment 5

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/0fe11ce43945

Comment 6

[Botond Ballo [:botond]]

Thanks again, Michael!

In case you're interested in working on something a bit more involved in the same part of the codebase, bug 1199798 is concerned with changing some of the other tree traversals in Async Pan/Zoom code to use BreadthFirstSearch() / DepthFirstSearch() instead of a manual loop, and extending these functions to serve the needs of a greater variety of call sites. It's a pretty open-ended bug which doesn't need to be done all at once - you could do however much of it you feel like.