STR: 1. Download https://dl.dropboxusercontent.com/u/40949268/emcc/TappyChickenUE4/TappyChicken.zip 2. Unzip, and run TappyChicken.html in the browser. 3. Set the pref devtools.performance.memory.sample-probability;1 to ensure that all allocations are captured and none are lost. 4. Enable "Record Memory" in the devtools profiler, and start profiling when in the main menu in the game. With profiling is active, click the Play button, and play for ~15 seconds at least, then stop profiling. Observed: a. I understand that Total Sampled Allocations should mean the number of allocations in the given node, plus all the children. However in the generated profile, it is seen that a child node has a larger Total Sampled Allocations count than a parent, which shouldn't be possible. b. There are multiple leaf nodes in the tree with 0 Total + Self Sampled Allocations, which shouldn't occur. See the attached screenshot that illustrates these issues.

Once bug 1172180 lands, allocations will be able to be used again, and there've been a few changes in how the allocations tree view gets rendered, so hoping this is no longer an issue, or if there is, I'd imagine it's the foundation of how this data is displayed at all. Jukka, if you can try the patch here, and see if this data is still strange to you, let me know! https://bug1172180.bmoattachments.org/attachment.cgi?id=8642550

Jordan, once that lands, can you also land a regression test for the view layer here? Something that just tests that we do in fact get allocations rendered in the tree view when we record allocations and allocate within that recording. Apparently unit + RDP level tests was not enough; I'd really like to avoid this again.

Yes -- bug 1177558 will handle the testing component to ensure the tree is not empty

I think this is fixed, and tested now that we had time to give this feature some love. Please file a new bug if you notice anymore weirdness. Thanks!