)]}'
{
  "commit": "83eafa8ca679f06f0364434bdb11bd3e6dca034f",
  "tree": "8fa26223eac1f3a6771966d47bbffdce1e759c68",
  "parents": [
    "148fcb3c7998713a5e6b239f526e9194c8905ad7"
  ],
  "author": {
    "name": "Behdad Esfahbod",
    "email": "behdad@behdad.org",
    "time": "Thu Aug 27 16:23:30 2026 -0600"
  },
  "committer": {
    "name": "Behdad Esfahbod",
    "email": "behdad@behdad.org",
    "time": "Thu Aug 27 16:34:19 2026 -0600"
  },
  "message": "[CFF] Bound charstring operators by the live work budget\n\nThe per-operator budget spend added a check to the charstring\ninterpreter\u0027s innermost loop.  The loop already counts operators through\nmax_ops, so cap max_ops by the caller\u0027s remaining budget and charge the\nconsumed operators back after the loop.  A single counter then enforces\nboth the structural HB_CFF_MAX_OPS limit and the live work budget, with\nno budget check in the hot loop, and the operator count can never\novershoot the remaining budget.\n\nThis bounds the real quantity a COLR or VARC driver multiplies: executed\noperator count, which a small charstring can drive up to HB_CFF_MAX_OPS\nvia subroutine reuse regardless of its byte length -- so charging the\ncharstring\u0027s length instead would badly undercount.  CFF2 blend still\ncharges its proportional work in place.\n\nTests: meson test -C build (275/275 passed); the CFF/COLR gpu-fuzzer\nregression seed completes in 0.45s.\n\nAssisted-by: Claude \u003cnoreply@anthropic.com\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "a07939229e18832738dc0df7910dab89a38f7d0a",
      "old_mode": 33188,
      "old_path": "src/hb-cff-interp-cs-common.hh",
      "new_id": "625519f07559d49ee36838d28426d4f30b9477b2",
      "new_mode": 33188,
      "new_path": "src/hb-cff-interp-cs-common.hh"
    }
  ]
}