git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH 0/7] safer "push --force" with compare-and-swap

From
Junio C Hamano <gitster@pobox.com>
Date
Jul 9, 2013, 19:53 UTC
Message-ID
<1373399610-8588-1-git-send-email-gitster@pobox.com>
In-Reply-To
<7vfvvwk7ce.fsf@alter.siamese.dyndns.org>

When you have to replace an already published branch with its rebased version, you would have to --force, but then you risk losing work, if any, that was pushed by somebody else while you are working on rebasing, as your earlier decision that replacing the old one with its rebase is OK was based on the assumption that there is nothing else going on. Unfortunately, --force is blind, and did not offer this "... but fail if the tip has moved from what I expect".

And here is a series to remedy it. It lets you specify what the expected current values of refs you are attempting to update, and if you are lazy, the value is taken from your remote tracking branch for the ref you are attempting to update.

I am not married to the "lockref" name, but I think the semantics implemented here is what we discussed to do in the earlier discussion.

cf. http://thread.gmane.org/gmane.comp.version-control.git/229430
This may still be rough at edges, but a basic testset seems to pass.

I haven't bothered to check the smart-http transport; help is greatly appreciated.

Junio C Hamano (7):
  cache.h: move remote/connect API out of it
  builtin/push.c: use OPT_BOOL, not OPT_BOOLEAN
  push: beginning of compare-and-swap "force/delete safety"
  remote.c: add command line option parser for --lockref
  push --lockref: implement logic to populate old_sha1_expect[]
  t5533: test "push --lockref"
  push: document --lockref
 Documentation/git-push.txt |  26 +++++++
 builtin/fetch-pack.c       |   2 +
 builtin/push.c             |  18 ++++-
 builtin/receive-pack.c     |   1 +
 builtin/send-pack.c        |  26 +++++++
 cache.h                    |  62 ----------------
 connect.c                  |   1 +
 connect.h                  |  13 ++++
 fetch-pack.c               |   1 +
 fetch-pack.h               |   1 +
 refs.c                     |   8 ---
 remote.c                   | 149 +++++++++++++++++++++++++++++++++++++-
 remote.h                   |  83 +++++++++++++++++++++
 send-pack.c                |   2 +
 t/t5533-push-cas.sh        | 176 +++++++++++++++++++++++++++++++++++++++++++++
 transport-helper.c         |   6 ++
 transport.c                |  13 ++++
 transport.h                |   5 ++
 upload-pack.c              |   1 +
 19 files changed, 519 insertions(+), 75 deletions(-)
 create mode 100644 connect.h
 create mode 100755 t/t5533-push-cas.sh
-- 
1.8.3.2-875-g76c723c
Previous: Junio C HamanoNext: Junio C Hamano
Message 16 of 62 in “[RFD] Making "git push [--force/--delete]" safer?”
  1. Junio C HamanoJul 2, 2013
  2. Johan HerlandJul 2, 2013
  3. Johan HerlandJul 3, 2013
  4. Junio C HamanoJul 3, 2013
  5. Johan HerlandJul 3, 2013
  6. Jonathan del StrotherJul 3, 2013
  7. Johan HerlandJul 3, 2013
  8. Michael HaggertyJul 3, 2013
  9. Johannes SixtJul 3, 2013
  10. Junio C HamanoJul 3, 2013
  11. Johannes SixtJul 4, 2013
  12. Junio C HamanoJul 4, 2013
  13. Junio C HamanoJul 3, 2013
  14. Junio C HamanoJul 3, 2013
  15. Junio C HamanoJul 3, 2013
  16. 0/7 safer "push --force" with compare-and-swapJunio C Hamano, Jul 9, 2013
  17. 1/7 cache.h: move remote/connect API out of itJunio C Hamano, Jul 9, 2013
  18. 2/7 builtin/push.c: use OPT_BOOL, not OPT_BOOLEANJunio C Hamano, Jul 9, 2013
  19. 3/7 push: beginning of compare-and-swap "force/delete safety"Junio C Hamano, Jul 9, 2013
  20. 4/7 remote.c: add command line option parser for --lockrefJunio C Hamano, Jul 9, 2013
  21. John KeepingJul 16, 2013
  22. Junio C HamanoJul 17, 2013
  23. Junio C HamanoJul 17, 2013
  24. 5/7 push --lockref: implement logic to populate old_sha1_expect[]Junio C Hamano, Jul 9, 2013
  25. 6/7 t5533: test "push --lockref"Junio C Hamano, Jul 9, 2013
  26. 7/7 push: document --lockrefJunio C Hamano, Jul 9, 2013
  27. Aaron SchrabJul 9, 2013
  28. Junio C HamanoJul 9, 2013
  29. Johannes SixtJul 9, 2013
  30. Junio C HamanoJul 9, 2013
  31. Johannes SixtJul 9, 2013
  32. Junio C HamanoJul 9, 2013
  33. Junio C HamanoJul 9, 2013
  34. Johannes SixtJul 11, 2013
  35. Junio C HamanoJul 11, 2013
  36. Junio C HamanoJul 11, 2013
  37. Johannes SixtJul 12, 2013
  38. Junio C HamanoJul 12, 2013
  39. Johannes SixtJul 12, 2013
  40. Junio C HamanoJul 12, 2013
  41. Johannes SixtJul 13, 2013
  42. Junio C HamanoJul 13, 2013
  43. Junio C HamanoJul 13, 2013
  44. Johannes SixtJul 13, 2013
  45. John KeepingJul 14, 2013
  46. Johannes SixtJul 13, 2013
  47. Junio C HamanoJul 14, 2013
  48. Johannes SixtJul 14, 2013
  49. Jonathan NiederJul 14, 2013
  50. Jonathan NiederJul 14, 2013
  51. Johannes SixtJul 14, 2013
  52. Jonathan NiederJul 14, 2013
  53. Junio C HamanoJul 15, 2013
  54. Jonathan NiederJul 15, 2013
  55. Junio C HamanoJul 15, 2013
  56. Johannes SixtJul 15, 2013
  57. Junio C HamanoJul 15, 2013
  58. Default expectation of --lockrefJunio C Hamano, Jul 15, 2013
  59. Johannes SixtJul 15, 2013
  60. Marc BranchaudJul 9, 2013
  61. Michael HaggertyJul 9, 2013
  62. Junio C HamanoJul 9, 2013

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.