メモリ関係は問題が発生した際に関連性を検証するのが難しい場合が多いので、余計な問題に遭遇しないように修正版を使用した方が賢明だと思います。

気になったもの:
SHA-1: 2496636a6536fcc19b0b0d8ad510ae9c64b424d7
* GPU: replace callocs with malloc since reallocs aren't cleared

SHA-1: c82371fc06ffb8d2970c985b44167e9a5e0222f1
* Mempool: simplify memory chunk list building

SHA-1: 412826a504a95c45b561893aa9cbaebd00d63702
* Mempool: delay allocating an initial chunk, its not always used

Crash関係:
SHA-1: c777f691ad427260fb39c8d191d3041169e557dd
* Fix T39663: Blender crashes when trying to triangulate uv unwrapped mesh.
　A dummy arg inversion! This is to be included in 2.70a!

SHA-1: ccf9afddbaa0a08f4e277b3c19bc15735be9f4a4
* Fix T39608: Blender 2.70 crashes when performing union
　This was a nasty bug which was caused by specific of how face-edge
　attributes are stored in Carve.
　Face pointer is used in the map key which works just fine in all
　cases except for the cases when some face is getting freed after
　it was stored in the map.
　This might give real issues when new face is allocating because
　it's possible new face would have the same address as the freed
　one.
　Such cases used to happen when union of separate manifolds is
　needed for the operands AND jemalloc is enabled.
　Solved by dropping attributes for the freed faces from the map.
　Maybe not the fastest ever approach, but not sure how to make
　it faster actually. Should work just fine. It only happens for
　complex setups with intersecting manifolds in the operands.

Crash関係:
SHA-1: 19413644dac0f56623c520c5da2039ed0294c220
* BGE: Fixing a crash when animating objects with modifiers and armatures.
　Our deformer system really needs some work. First, there was a crash
　with shape keys because BL_ModifierDeformer derives from
　BL_ShapeDeformer, which means we try to execute shape keys even if we do
　not have them. Also, for some reason BL_ModifierDeformer::Update() does
　not work if called from the threaded loop, so it is skipped for now. In
　other words, skinned updates on meshes with modifiers are currently not
　run in parallel.

気になったもの:
SHA-1: 5d63f162d596d34998b3cdb4733a7a228fcb3341
* Fix numpad emulation in non-US keyboards
　Patch D455 from Benoît Legat with own minor edits.

SHA-1: f700a13eb54c374b6df3c6494e7a10ca9991722c
* Revert "Mempool: simplify memory chunk list building"
　This reverts commit c82371fc06ffb8d2970c985b44167e9a5e0222f1.
　Caused regression in iterator

Crash関係:
SHA-1: 6300d22dabaec5dfa82fd290dc6a3036797507f8
* Fix T39640: Crash on maximizing/minimizing UV/Image Editor
　Added a NULL-pointer check for now, actual issue might be
　burried somewhere else (aka maybe traversal of the WM is
　not actually correct here?)

気になったもの:
SHA-1: b7b4f94e782f7e74c5acf0af19dfda07fdd92568
* Fix for Freestyle Python API modules not found in the Python Console.
　Addition of the path to the Freestyle Python API modules to 'sys.path' was delayed until
　the first Freestyle rendering, so that any import attempt of the modules in the Python
　Console always failed. Now the update of 'sys.path' is done at Blender start-up.
　This allows the Freestyle-specific modules to be imported without running Freestyle,
　facilitating quick interactive testing in the Console.

午前中、Gitから最新版を取得しようとするとエラーを吐いていたので明日にしようかと思ったが復旧したようだ。

Crash関係:
SHA-1: 7c78bcef6cb9c7cde928ed8062a12ab6ba88d2cb
* Fix for crash un-indenting in the text editor

Crash関係:
SHA-1: 3cf39fc72a36bab1c5b82ae96cffe7dd01ce7260
* Fix T39740: New Autosmooth instant crash entering Edit Mode
　One should always be wary of last minute changes, especially in comples code area... :/

気になったもの:
私はSCONSを使用しているので・・・私には関係ないですが。
SHA-1: d0c1b44debc4d4d1de274a4ecf777f76831cd264
* CMake: enable OpenMP for all platforms by default
　This is done in SCons and the only reason to disable OpenMP on
　Windows was to deal with the express version of compilers.
　Now when we are about to switch to msvc2013 by default we can
　safely enable OpenMP there since even express version of this
　beast supports OpenMP.

ビルドしたらPythonのランタイムエラーが出てビルドが強制終了されたけど・・・。
再度ビルドしたら正常にビルド完了した・・。
何が起こったのか・・意味不明・・奇々怪々。

Crash関係:
SHA-1: a15ae564217f2b47848c337e477068be0e150973
* Fix T39743: Crash when deleting faces in with new autosmooth.
　Odd I did not catch this one... :/

SHA-1: 97881d06b221fbe5db98c5e5b3d2b9ecd0a22b38
* Fix T39610: Shared mesh used for Mesh Deform causes crash
　For now disable using linked edit mesh in the meshdeform modifier.
　This is because editbmesh_get_derived_cage_and_final() might easily
　conflict with the thread which evaluates object which is in the edit
　mode for this mesh.
　We'll support this case once granular dependency graph is landed.

SHA-1: 4f1a5192c24595798942b6ce8d704031e9fda8de
* Fix T39742: Crash with Cycles + new autosmooth crash
　Nice little mistake, since the invalid mem access only happened once (the first time),
　was close to valid mem, and was only used to read, it would not crash often...

気になったもの:
SHA-1: d2a5ddb4ec12245b4d1580490d88e808a2a04761
* Optimisations for building "Long Keyframes"
　For a long time, one of the bottlenecks when drawing summary channels in the dopesheet
　(especially with many objects) was how the long keyframes feature (i.e showing holds
　between keyframes) got built. Specifically, it was the step where we check on the previous
　keyframe to see whether there's a hold between those two.
　The old code performed some elaborate checks, which made sense back when we used to handle
　certain summary channels (e.g. object-action/ipo, and groups IIRC) differently. However,
　nowadays, everything just does it by going over the FCurves one by one, so the offending
　code wasn't really providing much benefit. Unless I've forgotten some other reason why
　that old method is necessary, this commit should provide a decent speedup here, making
　things somewhat interactive now (if still a bit jerky).
　Other Tweaks:
　1) Introduced float-precision threshold when checking to see whether an existing long
　 keyframe could be reused. This should hopefully reduce the number of fp-jitter issues
　 when creating summaries for many channels, reducing the number of duplicates created.
　2) Precompute colours used for shading the long keyframes, instead of recomputing for
　 each block.

以下は個人的な印象なので無視してもらって構いません。
読み難くするために視認し難い色にしてあります。



システム開発に従事して来た者として、いつも思うことだがバグに関する対応が甘い様な気がする。
要件確認された成果物に対して不具合が見つかった場合それは確認項目に問題があると判断し項目自体の見直しから行われるべきで
対応が的確であると判断されるにはより多くの確認が成されるべきである。
しかし報告された箇所の検証がされただけで適用されているようにしか思えない・・・。
それ自体的に問題が内在している場合も過去に多々あったし・・・。
商業ベースの物との差と言えば・・・・・そうかもしれないが・・。
瑕疵認識が低いのは明確にドキュメント化されていないのが大きな要因だとは思う。
しかし波及効果は考慮されるべきだとも思うのだが・・。
個別事項に多くの時間を割けない現状もあるのだろうが・・。
しかし検証に時間を割き不具合を減らしてリリースするのが結果的には自分達の作業量の軽減に繋がるはずなのだが・・。
最終的な完成形が明確ではなく流動的なシステムの性かもしれないなぁ・・。

Crash関係:
微妙にCrashとは意味合いが異なるかもしれないが・・・
SHA-1: d050577176366617bcfdc86f9dd659143e024648
* Fix T39445: Async LibLoad Crash
　There was some deadlock due to trying manage Python's GIL. Instead of
　continuing to fight with it, anything needing to call into Python while
　conversion during lib loading is just delayed until it can be done in
　the main thread.

SHA-1: 70f089b6f5b768b7ee0d6bc98d64a202e99fc22f
* Fix T39509: Crash when assign an empty as a LoD
　Missing check when looking for appropriate LoD

FreeStyleに関する修正が多い。
また修正が間違っていて再度修正されていたりするので出来れば最新版を入手した方がいいかも知れない。

Crash関係:
SHA-1: ddc88ea419ef81d0a12900a71f183ac8298c3931
* Fix T39765 crash on text with generative modifiers.

SHA-1: c4a9ec88c8c7823f5a25aa28d21b4dc5c7d946ef
* Fix crash when enabling undistorted display in MCE

SHA-1: 88ae0be1c41bde777577cd9b08c6b963fbca7e0c
* Fix T39758: Blender Crash when removing pinned Scene.

SHA-1: ba61992cc3a98bdcc9cebd59ea96cfad5a63e08d
* Fix T39767: Parent in Mask mode crashes

気になったもの:
SHA-1: ed2ddc9f706b956ea955ac86b3e7ec5e0b41d9cd
* Support multiple distortion models, including a new division model
　This commit makes it so CameraIntrinsics is no longer hardcoded
　to use the traditional polynomial radial distortion model. Currently
　the distortion code has generic logic which is shared between
　different distortion models, but had no other models until now.
　This moves everything specific to the polynomial radial distortion
　to a subclass PolynomialDistortionCameraIntrinsics(), and adds a
　new division distortion model suitable for cameras such as the
　GoPro which have much stronger distortion due to their fisheye lens.
　This also cleans up the internal API of CameraIntrinsics to make
　it easier to understand and reduces old C-style code.
　New distortion model is available in the Lens panel of MCE.
　　- Polynomial is the old well-known model
　　- Division is the new one which s intended to deal better with huge
　　 distortion.
　Coefficients of this model works independent from each other
　and for division model one probably want to have positive values
　to have a barrel distortion.

SHA-1: bd3231c7604a346bbbb6adeb1d027077392eabe2
* Fix T39766: bone calculation from edit mode goes wrong
　Own logic error in previous commit...

Crash関係:
SHA-1: 0e2f6c7fc4db06b8fa9be758e294a3376c32bc5d
* Fix T39833: Nurbs Crash
　Converting empty object to mesh can generate a 'Null' BL:Mesh, we have to check against it.