Skip to content

Improvements for ref storage formats with submodules

Patrick Steinhardt git 2 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v2 Latest8 patches 3 repliesIn review
  2. v16 patches 10 repliesSuperseded

Changes in v2 from the cover letter

- I noticed that git submodule add also used to fail when used with a preexisting repository that uses a different ref format.

- Add a preparatory refactoring that wraps overly long command lines in "git-submodule.sh".

- Wire up support for --ref-format= in git submodule add.

Patches in v2 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/8git-submodule.sh: break overly long command linesgit-submodule.sh+55 -90 replies
2/8builtin/submodule: allow cloning with different ref storage formatDocumentation/git-submodule.txt builtin/submodule--helper.c git-submodule.sh +1 more+84 -10 replies
3/8builtin/clone: propagate ref storage format to submodulesbuiltin/clone.c t/t7424-submodule-mixed-ref-formats.sh+31 -22 replies
4/8refs: fix ref storage format for submodule ref storesrefs.c t/t7424-submodule-mixed-ref-formats.sh+70 -20 replies
5/8builtin/submodule: allow "add" to use different ref storage formatDocumentation/git-submodule.txt builtin/submodule--helper.c git-submodule.sh +1 more+39 -20 replies
6/8submodule: fix leaking fetch taskssubmodule.c+7 -100 replies
7/8submodule: fix leaking seen submodule namessubmodule.c t/t5572-pull-submodule.sh t/t7418-submodule-sparse-gitmodules.sh+3 -00 replies
8/8object: fix leaking packfiles when closing object storeobject.c t/t7424-submodule-mixed-ref-formats.sh+10 -00 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v2 yet.

Read the v2 thread

Cover letter [PATCH v2 0/8] Improvements for ref storage formats with submodules

Hi,

this is the second version of my patch series that aims to improve interaction of ref storage formats with submodules.

Changes compared to v1:

- I noticed that git submodule add also used to fail when used with a preexisting repository that uses a different ref format.

- Add a preparatory refactoring that wraps overly long command lines in "git-submodule.sh".

- Wire up support for --ref-format= in git submodule add.

Notably not included is a change to make git submodule add/clone honor the ref storage format of the parent repository. I think it is somewhat orthogonal to the changes I do in this series, and it is not entirely obvious whether it is a good idea or not. We can still do this in a future patch series, if we ever hear good arguments for it.

Thanks!

Patrick Steinhardt (8):
git-submodule.sh: break overly long command lines builtin/submodule: allow cloning with different ref storage format builtin/clone: propagate ref storage format to submodules
refs: fix ref storage format for submodule ref stores builtin/submodule: allow "add" to use different ref storage format
submodule: fix leaking fetch tasks
submodule: fix leaking seen submodule names
object: fix leaking packfiles when closing object store

 Documentation/git-submodule.txt        |  10 +-
 builtin/clone.c                        |  10 +-
 builtin/submodule--helper.c            |  46 +++++++-
 git-submodule.sh                       |  82 ++++++++++++--
 object.c                               |   9 ++
 refs.c                                 |   2 +-
 submodule.c                            |  18 ++--
 t/t5572-pull-submodule.sh              |   1 +
 t/t7418-submodule-sparse-gitmodules.sh |   1 +
 t/t7424-submodule-mixed-ref-formats.sh | 144 +++++++++++++++++++++++++
 10 files changed, 298 insertions(+), 25 deletions(-)
 create mode 100755 t/t7424-submodule-mixed-ref-formats.sh

Range-diff against v1:

-:  ---------- > 1:  6513c6b17d git-submodule.sh: break overly long command lines
1:  a450759bd1 = 2:  e6cda43878 builtin/submodule: allow cloning with different ref storage format
2:  e5923c0b33 = 3:  ed314f5333 builtin/clone: propagate ref storage format to submodules
3:  aaff9134ed ! 4:  f13356581e refs: fix ref storage format for submodule ref stores

@@ Commit message
$ git pull --recursive
fatal: Unable to find current revision in submodule path 'path/to/sub'

- Fix the bug by using the submodule repository's ref storage format
- instead.
+ The same issue occurs when adding a repository contained in the working
+ tree with a different ref storage format via git submodule add.

- Note that only the second added test fails without this fix. The other
- one is included regardless as it exercises other parts where we might
- end up accessing submodule ref stores.
+ Fix the bug by using the submodule repository's ref storage format
+ instead and add some tests. Note that the test for `git submodule
+ status` was included as a precaution, only. The command worked alright
+ even without the bugfix.

@@ t/t7424-submodule-mixed-ref-formats.sh: do
+ # Some tests migrate the ref storage format, which does not work with
+ # reflogs at the time of writing these tests.
+ git config set --global core.logAllRefUpdates false
++'
++
++test_expect_success 'add existing repository with different ref storage format' '
++ test_when_finished "rm -rf parent" &&
++
++ git init parent &&
++ (
++ cd parent &&
++ test_commit parent &&
++ git init --ref-format=$OTHER_FORMAT submodule &&
++ test_commit -C submodule submodule &&
++ git submodule add ./submodule
++ )
'

test_expect_success 'recursive clone propagates ref storage format' '

-:  ---------- > 5:  4ce17e44a1 builtin/submodule: allow "add" to use different ref storage format
4:  8f8371c18a = 6:  d92770290f submodule: fix leaking fetch tasks
5:  732142aaa6 = 7:  e9421189ca submodule: fix leaking seen submodule names
6:  8dc7cc76d5 = 8:  d05737c75f object: fix leaking packfiles when closing object store
Signature · 1 line
2.46.0.46.g406f326d27.dirty