Hi,
this patch series is the result from the discussion at [1]. On the one
hand this series fixes the reported bug where dangling symrefs are not
renamed via `git remote rename`.
On the other hand this series reworks the logic used to rename remotes
so that we use two transactions instead of one transaction per ref. This
fixes quadratic runtime behaviour, where renaming 10k refs takes ~4
minutes, 100k takes hours. This results in a significant speedup with
both the "files" backend (benchmarked with a smaller number of refs to
retain sanity):
Benchmark 1: rename remote (refformat = files, revision = HEAD~)
Time (mean ± σ): 238.770 s ± 13.857 s [User: 91.473 s, System: 143.793 s]
Range (min … max): 204.863 s … 247.699 s 10 runs
Benchmark 2: rename remote (refformat = files, revision = HEAD)
Time (mean ± σ): 2.103 s ± 0.036 s [User: 0.360 s, System: 1.313 s]
Range (min … max): 2.011 s … 2.141 s 10 runs
Summary
rename remote (refformat = files, revision = HEAD) ran
113.53 ± 6.87 times faster than rename remote (refformat = files, revision = HEAD~)
For the "reftable" backend we see a significant speedup, as well, but
not as extreme as with the "files" backend:
Benchmark 1: rename remote (refformat = reftable, revision = HEAD~)
Time (mean ± σ): 8.604 s ± 0.539 s [User: 4.985 s, System: 2.368 s]
Range (min … max): 7.880 s … 9.556 s 10 runs
Benchmark 2: rename remote (refformat = reftable, revision = HEAD)
Time (mean ± σ): 1.177 s ± 0.103 s [User: 0.446 s, System: 0.270 s]
Range (min … max): 1.023 s … 1.410 s 10 runs
Summary
rename remote (refformat = reftable, revision = HEAD) ran
7.31 ± 0.79 times faster than rename remote (refformat = reftable, revision = HEAD~)
But in any case, it's one more case where the "reftable" backend
outperforms the "files" backend.
The series is built on top of e4ef0485fd7 (The fourteenth batch,
2025-07-24) with ps/reflog-migrate-fixes at de7cc0782a7 (refs: fix
invalid old object IDs when migrating reflogs, 2025-07-25) merged into
it.
I'd normally have withheld sending until that series was merged to
"next", but given that I promised to send something on Friday already I
decided to just get it out. In any case, if that causes problems I'm
happy to wait a bit before this series here gets merged into "seen".
Thanks!
Patrick
[1]: <CANrWfmQWa=RJnm7d3C7ogRX6Tth2eeuGwvwrNmzS2gr+eP0OpA@xxxxxxxxxxxxxx>
---
Patrick Steinhardt (4):
refs: pass refname when invoking reflog entry callback
refs: simplify logic when migrating reflog entries
builtin/remote: rework how remote refs get renamed
builtin/remote: only iterate through refs that are to be renamed
builtin/fsck.c | 9 +-
builtin/gc.c | 3 +-
builtin/remote.c | 275 ++++++++++++++++++++++++++++++----------------
builtin/stash.c | 6 +-
commit.c | 3 +-
object-name.c | 3 +-
reflog-walk.c | 7 +-
reflog.c | 3 +-
reflog.h | 3 +-
refs.c | 63 +++++------
refs.h | 3 +
