Re: [PATCH] merge: use repo_in_merge_bases for octopus up-to-date check
- From
Kristofer Karlsson <krka@spotify.com>
- Date
- May 18, 2026, 14:38 UTC
- Message-ID
- <CAL71e4NosWg_UwZ6fn0FuaTS89U6Sm9PWAx=gTjFMzMCsEOw6w@mail.gmail.com>
- In-Reply-To
- <c5b333f1-0db6-4aec-a369-6503cb924e7f@gmail.com>
Good question! No, this was intended only as a code cleanup and semantic simplification. The code path seems like an edge case, so benchmarking it did not seem worthwhile. The main win as I see it is clearer semantics for what it's doing and the optimization is just a bonus.
Checking if repo_get_merge_bases(HEAD, J) == J is better expressed as repo_in_merge_bases(J, HEAD)
repo_in_merge_bases should probably be renamed to something like repo_is_ancestor, since its comment says: Is "commit" an ancestor of (i.e. reachable from) the "reference"? but I can understand that it may be painful to rename it in practice.
If I do a mental rename, this becomes simpler to reason about:
Checking if repo_get_merge_bases(HEAD, J) == J is better expressed as repo_is_ancestor(J, HEAD)
- Kristofer
On Mon, 18 May 2026 at 14:20, Derrick Stolee <stolee@gmail.com> wrote:
Show 49 quoted lines
> > On 5/12/2026 2:11 AM, Kristofer Karlsson via GitGitGadget wrote: > > From: Kristofer Karlsson <krka@spotify.com> > > > > The octopus merge path checks whether each remote head is already > > an ancestor of HEAD by computing all merge-bases via > > repo_get_merge_bases() and comparing the first result's OID to > > the remote head. This is more expensive than necessary: > > repo_get_merge_bases() calls paint_down_to_common() with > > min_generation=0, performs the full STALE drain, and may run > > remove_redundant(), when all we need is a yes/no reachability > > answer. > > > > Replace this with repo_in_merge_bases(), which answers the > > is-ancestor question directly. When generation numbers are > > available, repo_in_merge_bases() uses can_all_from_reach() -- a > > DFS bounded by generation number that stops as soon as the target > > is found or ruled out, without entering paint_down_to_common() at > > all. Without generation numbers, it still benefits from a tighter > > min_generation floor. > > > > Signed-off-by: Kristofer Karlsson <krka@spotify.com> > > --- > > merge: use repo_in_merge_bases for octopus up-to-date check > > > > While reviewing callers of repo_get_merge_bases() for a different patch, > > I noticed the octopus up-to-date loop in builtin/merge.c computes full > > merge-bases only to check whether each remote head is an ancestor of > > HEAD. > > > > The existing code calls repo_get_merge_bases(), takes the first result, > > frees the list, and compares the OID to the remote head. This is > > equivalent to an is-ancestor check, which repo_in_merge_bases() answers > > directly. > > > > Using repo_in_merge_bases() simplifies the code (-14/+4 lines) and > > avoids unnecessary work: with generation numbers it uses > > can_all_from_reach() instead of paint_down_to_common(), and without > > generation numbers it still benefits from a tighter min_generation > > floor. In practice this only matters for octopus merges on repos with > > deep history, so the main value here is the simplification. > > The code change looks right to me. Do you have any performance numbers > to share? Or was this motivated mostly as an opportunity for code cleanup? > > Thanks, > -Stolee > >