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

[PATCH 00/14] Teach "git merge FETCH_HEAD" octopus merges

From
Junio C Hamano <gitster@pobox.com>
Date
Apr 26, 2015, 05:25 UTC
Message-ID
<1430025967-24479-1-git-send-email-gitster@pobox.com>
In-Reply-To
<xmqqiocqli1c.fsf@gitster.dls.corp.google.com>
Junio C Hamano <gitster@pobox.com> writes:
Show 17 quoted lines
> Jeff King <peff@peff.net> writes:
>> On Mon, Apr 20, 2015 at 11:59:04AM -0700, Junio C Hamano wrote:
>>
>>> Unfortunately, "git merge"'s parsing of FETCH_HEAD forgets that we
>>> may be creating an Octopus.  Otherwise the above should work well.
>>
>> That sounds like a bug we should fix regardless.
>
> But I am not sure how it should behave. "git fetch $there A B C"
> followed by "git merge FETCH_HEAD" merges only A, and I do not know
> if people have come to depend on this behaviour.
>
> I suspect there may be larger fallout from such a change, namely,
> what should "git log FETCH_HEAD" do?  Should it traverse the history
> starting from all things that are not marked as not-for-merge, or
> should it just say "git rev-parse FETCH_HEAD" and use only the first
> one as the starting point?
So, I thought we may want to try this and see how it goes.

Tentatively, I am saying that "FETCH_HEAD" is a magic token understood by "git merge", like "git merge <msg> HEAD commits..." syntax was a magic that made "git merge" work differently from "git merge -m <msg> <commits>..."; no changes to get_sha1() or anything heavy like that is intended.

Earlier, we thought that it would just be the matter of turning existing invocation of "git merge <msg> HEAD $commits..." in "git pull" into "git merge -m <msg> $commits..." to deprecate the ugly original "merge" command line syntax.

This unfortunately failed for two reasons.
 * "-m <msg>" stops the editor from running; recent "git pull"
   encourage the users to justify the merge in the log message,
   and the auto-generated <msg> that comes from format-merge-msg
   should still be shown to the user in the editor to be edited.
 * "-m <msg>" also adds auto-generated summary when merge.log
   configuration is enabled, but "git pull" calls "git merge" with
   the message _with_ that log already in there.

Invoking "git merge FETCH_HEAD" (no messages fed by the caller) from "git pull" almost works. "git merge" knows how to extract the name of the repository and the branch from FETCH_HEAD to use in the merge log message it auto-generates. However, this is done only for a single branch; if you did "git pull $there topic-A topic-B", and then invoked "git merge FETCH_HEAD" from there, we would end up recording a merge with only one branch, which is not what we want.

This teaches "git merge FETCH_HEAD" that FETCH_HEAD may describe multiple branches that were fetched for merging. With that, the last step eradicates the "merge <msg> HEAD <commit>..." syntax from our codebase, finally.

This may be rough in a few places and some patches that are done as preparatory clean-up steps may want to be squashed into the patch that follows them that implements the real change.

These patches are designed to apply on top of v2.2.2; I'll push them out on 'pu' later, on 'jc/merge' topic.

Junio C Hamano (14):
  merge: simplify code flow
  t5520: style fixes
  t5520: test pulling an octopus into an unborn branch
  merge: clarify "pulling into void" special case
  merge: do not check argc to determine number of remote heads
  merge: small leakfix and code simplification
  merge: clarify collect_parents() logic
  merge: split reduce_parents() out of collect_parents()
  merge: narrow scope of merge_names
  merge: extract prepare_merge_message() logic out
  merge: make collect_parents() auto-generate the merge message
  merge: decide if we auto-generate the message early in
    collect_parents()
  merge: handle FETCH_HEAD internally
  merge: deprecate 'git merge <message> HEAD <commit>' syntax
 Documentation/git-merge.txt   |   4 +
 builtin/merge.c               | 248 +++++++++++++++++++++++++++---------------
 git-cvsimport.perl            |   2 +-
 git-pull.sh                   |   3 +-
 t/t3402-rebase-merge.sh       |   2 +-
 t/t5520-pull.sh               |  31 +++---
 t/t6020-merge-df.sh           |   2 +-
 t/t6021-merge-criss-cross.sh  |   6 +-
 t/t9402-git-cvsserver-refs.sh |   2 +-
 9 files changed, 188 insertions(+), 112 deletions(-)
-- 
2.4.0-rc3-238-g36f5934
Previous: Junio C HamanoNext: Junio C Hamano
Message 8 of 42 in “[BUG] "git pull" will regress between 'master' and 'pu'”
  1. Junio C HamanoApr 19, 2015
  2. Jeff KingApr 19, 2015
  3. brian m. carlsonApr 19, 2015
  4. Jeff KingApr 19, 2015
  5. Junio C HamanoApr 20, 2015
  6. Jeff KingApr 20, 2015
  7. Junio C HamanoApr 20, 2015
  8. 00/14 Teach "git merge FETCH_HEAD" octopus mergesJunio C Hamano, Apr 26, 2015
  9. 01/14 merge: simplify code flowJunio C Hamano, Apr 26, 2015
  10. 02/14 t5520: style fixesJunio C Hamano, Apr 26, 2015
  11. 03/14 t5520: test pulling an octopus into an unborn branchJunio C Hamano, Apr 26, 2015
  12. 04/14 merge: clarify "pulling into void" special caseJunio C Hamano, Apr 26, 2015
  13. 05/14 merge: do not check argc to determine number of remote headsJunio C Hamano, Apr 26, 2015
  14. 06/14 merge: small leakfix and code simplificationJunio C Hamano, Apr 26, 2015
  15. 07/14 merge: clarify collect_parents() logicJunio C Hamano, Apr 26, 2015
  16. 08/14 merge: split reduce_parents() out of collect_parents()Junio C Hamano, Apr 26, 2015
  17. 09/14 merge: narrow scope of merge_namesJunio C Hamano, Apr 26, 2015
  18. 10/14 merge: extract prepare_merge_message() logic outJunio C Hamano, Apr 26, 2015
  19. 11/14 merge: make collect_parents() auto-generate the merge messageJunio C Hamano, Apr 26, 2015
  20. 12/14 merge: decide if we auto-generate the message early in collect_parents()Junio C Hamano, Apr 26, 2015
  21. 13/14 merge: handle FETCH_HEAD internallyJunio C Hamano, Apr 26, 2015
  22. 14/14 merge: deprecate 'git merge <message> HEAD <commit>' syntaxJunio C Hamano, Apr 26, 2015
  23. 00/15 Teach "git merge FETCH_HEAD" octopus mergesJunio C Hamano, Apr 29, 2015
  24. 01/15 merge: test the top-level merge driverJunio C Hamano, Apr 29, 2015
  25. 02/15 merge: simplify code flowJunio C Hamano, Apr 29, 2015
  26. 03/15 t5520: style fixesJunio C Hamano, Apr 29, 2015
  27. Paul TanMay 1, 2015
  28. Junio C HamanoMay 3, 2015
  29. 04/15 t5520: test pulling an octopus into an unborn branchJunio C Hamano, Apr 29, 2015
  30. 05/15 merge: clarify "pulling into void" special caseJunio C Hamano, Apr 29, 2015
  31. 06/15 merge: do not check argc to determine number of remote headsJunio C Hamano, Apr 29, 2015
  32. 07/15 merge: small leakfix and code simplificationJunio C Hamano, Apr 29, 2015
  33. 08/15 merge: clarify collect_parents() logicJunio C Hamano, Apr 29, 2015
  34. 09/15 merge: split reduce_parents() out of collect_parents()Junio C Hamano, Apr 29, 2015
  35. 10/15 merge: narrow scope of merge_namesJunio C Hamano, Apr 29, 2015
  36. 11/15 merge: extract prepare_merge_message() logic outJunio C Hamano, Apr 29, 2015
  37. 12/15 merge: make collect_parents() auto-generate the merge messageJunio C Hamano, Apr 29, 2015
  38. 13/15 merge: decide if we auto-generate the message early in collect_parents()Junio C Hamano, Apr 29, 2015
  39. 14/15 merge: handle FETCH_HEAD internallyJunio C Hamano, Apr 29, 2015
  40. 15/15 merge: deprecate 'git merge <message> HEAD <commit>' syntaxJunio C Hamano, Apr 29, 2015
  41. Junio C HamanoApr 20, 2015
  42. Johannes SchindelinApr 21, 2015

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.