)]}'
{
  "commit": "441c439b9916474cac15d2822af47a9ad70674a8",
  "tree": "6ed54a8e493555263794d7a15e56213270e857f6",
  "parents": [
    "bf89bdb013bca4bd547cdbaf11153333e0ec332a",
    "de4c1d1edc49723a78954d30a83690aa1937422f"
  ],
  "author": {
    "name": "Wan-Teh Chang",
    "email": "wtc@google.com",
    "time": "Tue Aug 25 13:22:19 2026 -0700"
  },
  "committer": {
    "name": "Wan-Teh Chang",
    "email": "wtc@google.com",
    "time": "Tue Aug 25 13:22:19 2026 -0700"
  },
  "message": "Merge tag \u0027v3.15.0\u0027 into main branch\n\nRelease v3.15.0 Quibble Pants\n\n2026-08-25 v3.15.0\n  This release is ABI compatible with the last release.\n\n  This release includes the fixes for all the security issues found by\n  AI tools recently.\n\n  Important build system change: The cmake option ENABLE_APPS was added\n  to build aomdec and aomenc without building the example programs. The\n  cmake option ENABLE_EXAMPLES was redefined to build only the example\n  programs.\n\n  - New Features\n    * Test only: New codec control AV1E_SET_MODE_REF_DELTA_ENABLED for\n      the encoder to toggle loopfilter mode_ref_delta_enabled.\n    * New codec option “force-max-q” for the encoder to force the use\n      of max Q in VBR mode.\n    * Completed low-complexity decode mode development for VOD video,\n      and supported HDres (i.e. 720p to 1080p) encoding.\n    * IntraBC for RTC screen: improvements to allow and speed up this\n      tool for realtime mode.\n\n  - Compression Efficiency Improvements\n    * Improved the rate allocation among hierarchical layers and luma\n      and chroma planes.\n\n  - Perceptual Quality Improvements\n    * Improved VMAF metric while maintaining PSNR performance.\n\n  - Speedups\n    * Continued the speed 0 to 3 encoder speed optimizations.\n    * Improvements to Arm optimizations for convolve and variance\n      functions.\n\n  - Bug Fixes\n    * Bug 541725398: aom_img_flip: fix chroma height calculation\n    * Bug 437872594: CONFIG_SHARED\u003d1 produces an .a(rchive) file for\n      static linking\n    * Bug 503326506: SEGV in encoder_destroy via failed encoder\n      initialization\n    * Bug 538651949: Bounds-check bypass in AV1E_SET_SVC_LAYER_ID:\n      ppi/svc layer-count desync re-enables OOB layer_context[]\n      read/write (incomplete fix for issue 503993985 / CVE-2026-56209)\n    * Bug 525452409: The math and Threads::Threads library dependencies\n      should not be part of the link interface of the aom shared library\n    * Bug 524353636: AOM_EXTRA_C_FLAGS/-DSANITIZE\u003d flags not propagated\n      w/CMAKE_BUILD_TYPE\u003dDebug\n    * Bug 513342555: Potential heap double-free in libaom AV1 MT\n      encoding via aliased macroblock data\n    * Bug 520656240: Copy on write: libaom AV1 wedge/interintra mask\n      buffers could be precomputed\n    * Bug 514696186: aomdec fails with \"Failed to decode tile data\" on\n      stream encoded by aomenc\n    * Bug 382465458: Chessboard pattern on flat gray video encoded with\n      libaom AV1\n    * oss-fuzz:514006304: Increase ctx-\u003ecx_data_sz to 2.5x uncomp frame\n      size\n    * Bug 514816767: aomenc fails assertion check mi-\u003ebsize \u003d\u003d bsize\n    * Bug 504613867: Potential heap OOB read in\n      av1_int_pro_motion_estimation due to stale stride (CVE-2026-13906)\n    * Bug 503810640: Undefined behavior: NaN-to-int conversion in\n      aom_noise_model_get_grain_parameters\n    * Bug 503171639: Heap-buffer-overflow in\n      av1_update_and_record_txb_context via SVC encoding with mismatched\n      TXB buffer size\n    * Bug 505976409: SIGFPE (integer divide-by-zero) in libaom AV1 SVC\n      encoder\n\nBug: 549271249\nChange-Id: I4a10986ae5323fa5e593217a81803aaa0552cd4c\n",
  "tree_diff": []
}