• @[email protected]
    link
    fedilink
    1
    edit-2
    7 months ago

    You seem to be talking about binary search, but this is a search with an unbounded end.

    I think the actual optimal thing would be just to take the first commit and bisect from there. But in practice this has downsides:

    1. Bugs that you are tracking down are typically biased towards recent commits.
    2. Older commits may not even be relevant to the bug (ex: before the feature was introduced)
    3. You may have trouble building older commits for various reasons.

    So I think “optimal” is sort of fuzzy in this context, but I’m sure you could come up with various models for these variables and come up with something that is optimal for a given model. I haven’t got around to doing that yet though.

    • @cbarrick
      link
      English
      3
      edit-2
      7 months ago

      No, I’m talking about exponential search which is like binary search but unbounded end.

      OP has implemented an exponential search.