)]}'
{
  "commit": "819c1fd4c24ea59c1af6a205a5c82944c6fc96cd",
  "tree": "245024da884dfb185d05fbe1a397185e2c312eee",
  "parents": [
    "74c32399b100531b22184bd6ac38c4b26bc86ca5"
  ],
  "author": {
    "name": "Philipp Wollermann",
    "email": "philwo@google.com",
    "time": "Mon Sep 21 11:34:28 2026 +0200"
  },
  "committer": {
    "name": "gn-scoped@luci-project-accounts.iam.gserviceaccount.com",
    "email": "gn-scoped@luci-project-accounts.iam.gserviceaccount.com",
    "time": "Wed Nov 11 17:22:18 2026 -0800"
  },
  "message": "Header check: wait for chunked work with std::latch\n\nDoWorkChunked posts a known number of chunks to the worker pool, so a\nstd::latch initialized with the chunk count replaces the AtomicRefCount,\nmutex, and condition variable that tracked completion. This removes the\nextra reference held while posting and the lock held across the\ndecrement, both of which only existed to keep the completion state alive\nuntil the last notification finished.\n\nstd::latch::count_down only touches the latch address after the final\ndecrement, so destroying the latch as soon as wait() returns is safe.\nAll supported toolchains build GN as C++23 with a macOS minimum of 14,\nwhere std::latch is available.\n\nFollows up on a suggestion by msta@ on my earlier CL 26341. gn_unittests\nand the simple integration test pass, no measurable performance impact.\n\nChange-Id: I355fcb0c57716d3a66e79f3268886714c2aeef54\nReviewed-on: https://gn-review.googlesource.com/c/gn/+/26640\nCommit-Queue: Philipp Wollermann \u003cphilwo@google.com\u003e\nReviewed-by: Takuto Ikuta \u003ctikuta@google.com\u003e\nReviewed-by: Matt Stark \u003cmsta@google.com\u003e\n",
  "tree_diff": [
    {
      "type": "modify",
      "old_id": "69c2c95aa076a553ef0806f83d43296580dabea7",
      "old_mode": 33188,
      "old_path": "src/gn/header_checker.cc",
      "new_id": "6d2c279658b127130399f33cb53004357477e133",
      "new_mode": 33188,
      "new_path": "src/gn/header_checker.cc"
    }
  ]
}
