)]}'
{
  "commit": "e742f55bd2fc24cfba8575e4ebcaa64df1116d74",
  "tree": "575c2a50ff5f3ae9c249e4ea90a3bdc63595f094",
  "parents": [
    "16d437019d1c15ebe16ca4d5eeddbba674654064",
    "f2a332d7b80cfe06ac53fc63c6e7006396273298"
  ],
  "author": {
    "name": "Jonathan Brouwer",
    "email": "jonathantbrouwer@gmail.com",
    "time": "Tue Nov 10 21:42:01 2026 +0100"
  },
  "committer": {
    "name": "GitHub",
    "email": "noreply@github.com",
    "time": "Tue Nov 10 21:42:01 2026 +0100"
  },
  "message": "Rollup merge of #165611 - panstromek:no-terminator-option-4, r\u003doli-obk\n\nAvoid terminator: None in MIR building\n\nAnother step towards changing `Option\u003cTerminator\u003e` to `Terminator` in `BasicBlockData`. More info in [this thread](https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/vibe.20check.20-.20removing.20Option.20from.20Terminator/with/637588808).\n\nMIR building sometimes needs to keep blocks unterminated, so we model this explicitly outside of the `BasicBlockData` structure by using `Either\u003cBasicBlockData, UnterminatedBlockData\u003e`. Notably, this type has the same layout as `BasicBlockData`, so we don\u0027t need to reallocate when finalizing the MIR body.\n\nThe relevant perf run is [the first one](https://github.com/rust-lang/rust/pull/165611#issuecomment-6712547935), the improvement is noise.\n\nI also tried using custom block representation that would keep terminator optional in [the second perf run](https://github.com/rust-lang/rust/pull/165611#issuecomment-6720151494), but it was a regression, because the blocks have to be reallocated and copied when building is finished. See https://github.com/rust-lang/rust/pull/165611#issuecomment-6720831119 for the analysis.\n",
  "tree_diff": []
}
