)]}'
{
  "commit": "9d8f724069523041e2dbb3d848a621068a33a377",
  "tree": "5d49835e52cce1d32dc5477aac13027c644b4a55",
  "parents": [
    "bc65c019404b2f9b5a6894e42a30f1c398d6da99"
  ],
  "author": {
    "name": "Elliott Hughes",
    "email": "enh@google.com",
    "time": "Tue Aug 13 15:45:07 2024 +0000"
  },
  "committer": {
    "name": "Elliott Hughes",
    "email": "enh@google.com",
    "time": "Tue Aug 13 15:45:07 2024 +0000"
  },
  "message": "Use size_t for stdio\u0027s __sbuf even on ILP32.\n\nThis is probably more of a theoretical bug than a practical one, since allocating enough memory to get yourself into trouble in a 32-bit process is non-trivial, but there\u0027s no obvious reason for this LP64/ILP32 difference here.\n\nAt the time we ported bionic to LP64, this struct was still in \u003cstdio.h\u003e, which probably explains the reticence to just change ILP32 from int to size_t. Even though those types have the same size and alignment, and it\u0027s unlikely that anyone\u0027s code would be broken by a change of signedness, the same change changed the short that contains an fd in __sFILE to int, which definitely is an ABI change (that at the time at least was app-breaking, because apps were definitely poking about in __sFILE directly).\n\nChange-Id: I0b938fb6cb0da3a241a5ba4a4cda1b867a2578c1\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "62efea1ba263edc927e477c94b1f926d5491e89a",
      "old_mode": 33188,
      "old_path": "libc/stdio/local.h",
      "new_id": "a60468efaee4d35d2ef7be4638fc30d33192dacc",
      "new_mode": 33188,
      "new_path": "libc/stdio/local.h"
    }
  ]
}