This causes leaks, which we now catch in debug builds (thanks ehsan!)

Comment on attachment 8389451 [details] [diff] [review] Delete them! Review of attachment 8389451 [details] [diff] [review]: ----------------------------------------------------------------- Let's stop using mozilla::LinkedList, it's caused several problems already, the semantics are insane. Let's just use an STL container.

I went with nsTArray since policy is to stick with those classes unless we really need something specific to std:: (to maintain consistency across the code base). I think array/vector is better than list too here, since we're basically always going to have two pools, and we can allocate the size upfront.

Backed out along with bug 982318 in https://hg.mozilla.org/integration/mozilla-inbound/rev/892dd317097f for b2g mochitest crash/hangs (nobody can really tell which they do) like https://tbpl.mozilla.org/php/getParsedLog.php?id=36315626&tree=Mozilla-Inbound in most chunks.

Bug 985302 is necessary for this bug. After applying attachment 8389581 [details] [diff] [review] to b2g v1.4, I faced Bug 985302 problem.

A patch for b2g v1.4.

(In reply to Sotaro Ikeda [:sotaro] from comment #9) > Created attachment 8411320 [details] [diff] [review] > Use nsTArray v2 for b2g v1.4 > > A patch for b2g v1.4. Just unbitrot attachment 8389581 [details] [diff] [review].

Nominate to "1.4+", because this bug blocks Bug 998504.

Backed out for frequent Android crashes. https://hg.mozilla.org/releases/mozilla-aurora/rev/95ed6fad0a6c https://tbpl.mozilla.org/php/getParsedLog.php?id=38434306&tree=Mozilla-Aurora https://tbpl.mozilla.org/php/getParsedLog.php?id=38434603&tree=Mozilla-Aurora https://tbpl.mozilla.org/php/getParsedLog.php?id=38434728&tree=Mozilla-Aurora

B2G as well. https://tbpl.mozilla.org/php/getParsedLog.php?id=38435389&tree=Mozilla-Aurora

Sorry, I created the patch based on attachment 8389581 [details] [diff] [review]. It was wrong. After that there was an update to the patch and updated patch was checked in to m-c. The crash happened because of lacking a cancel timer in TextureClientPool::~TextureClientPool(). It is not present in attachment 8389581 [details] [diff] [review]. But it is added to m-c checked in version.

(In reply to Carsten Book [:Tomcat] from comment #7) > https://hg.mozilla.org/mozilla-central/rev/2a19ce01687f I confirmed that the above checked-in diff could be safely applied to b2g v1.4.

aurora specific patch is not necessary. m-c checked in diff could be uplifted to aurora.

It is necessary only when tiling is enable. Tiling is a capability that enabled since b2g v1.4

James, we have no tiling on v1.3t, so we don't need that.