)]}'
{
  "commit": "042e90c941026cd14d49ef853d3373d569ff6e36",
  "tree": "abf68f87ca58c8bbde4364243abc925fc0f8e02b",
  "parents": [
    "dd1659e492c36fdee7433abf4e0a823609643cd2"
  ],
  "author": {
    "name": "Jonathan Wright",
    "email": "jonathan.wright@arm.com",
    "time": "Wed Jan 18 00:32:45 2023 +0000"
  },
  "committer": {
    "name": "Wan-Teh Chang",
    "email": "wtc@google.com",
    "time": "Fri Jan 20 01:40:42 2023 +0000"
  },
  "message": "Fix buffer overrun in dist_wtd_convolve_2d_horiz_neon\n\nWhen introducing the 6-tap specialization for\ndist_wtd_convolve_2d_vert_neon[1], we attempted to reduce the number\nof rows processed in the horizontal convolution (by 2) if the\nsubsequent vertical convolution would be using a 6-tap filter instead\nof an 8-tap filter. This logic was faulty and meant we ended up\naccessing memory outside of the (padded) source buffer - since the\nhorizontal convolution processes 4 rows of data per iteration, when\nthe number of rows to be processed is not necessarily a multiple of\n4.\n\nThis patch restores the previous logic and src pointer starting\nposition for dist_wtd_convolve_2d_horiz_neon.\n\n[1] Commit hash: be1d80024928684b1c9eebc648ed92d8ea70d166\n\nBug: aomedia:3367\nChange-Id: Id11d4ae7c123957d3336b61f4d4cc8da09131b68\n(cherry picked from commit 81da208fb6e0a1fa6a3714a7b7dbe3f613e8bc64)\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "700dc546da76dc4525b72e2c3fe93cd98af7bef1",
      "old_mode": 33188,
      "old_path": "av1/common/arm/jnt_convolve_neon.c",
      "new_id": "36c8f9cb8e8ee9541202daf6dd064ec30235c7ed",
      "new_mode": 33188,
      "new_path": "av1/common/arm/jnt_convolve_neon.c"
    }
  ]
}