[v6] merge: avoid write merge state when unable to write index

Commit Message

Kyle Zhao June 17, 2024, 3:08 a.m. UTC

From: Kyle Zhao <[e-mail]>

Writing the merge state after the index write fails is meaningless and
could potentially cause Git to lose changes.

Signed-off-by: Kyle Zhao <[e-mail]>
---
    merge: avoid write merge state when unable to write index
    
    In some of our monorepos, code is sometimes lost after merging.
    
    After investigation, we discovered the problem.
    
    This happens if we perform "git pull" or "git merge" when another git
    process is writing to the index, especially in a monorepo (because its
    index will be larger).
    
    How to reproduce:
    
    git init demo
    cd demo
    touch 1.txt && git add . && git commit -m "1"
    git checkout -b source-branch
    touch 2.txt && git add . && git commit -m "2"
    git checkout master
    echo "1" >> 1.txt && git add . && git commit -m "3"
    # another git process runnning
    touch .git/index.lock
    git merge source-branch
    # another git process finished
    rm .git/index.lock
    git commit -m "4"
    
    
    Then the modifications from the source branch are lost.
    
    Regards, Kyle

Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1731%2Fkeyu98%2Fkz%2Ffix-merge-when-index-lock-exists-v6
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1731/keyu98/kz/fix-merge-when-index-lock-exists-v6
Pull-Request: https://github.com/gitgitgadget/git/pull/1731

Range-diff vs v5:

 1:  a5156088514 ! 1:  24150cd814a merge: avoid write merge state when unable to write index
     @@ Metadata
       ## Commit message ##
          merge: avoid write merge state when unable to write index
      
     -    When running a merge while the index is locked (presumably by another
     -    process), the merge state is written, the index is not updated, and then
     -    the merge fails. This might cause unexpected results.
     -
     -    E.g., if another running process is "git commit", MERGE_HEAD and other
     -    state files we write on our side will be taken into account by them and
     -    cause them to record a merge, even though they may have been trying to
     -    record something entirely different.
     +    Writing the merge state after the index write fails is meaningless and
     +    could potentially cause Git to lose changes.
      
          Signed-off-by: Kyle Zhao <[e-mail]>
      


 builtin/merge.c  |  2 +-
 t/t7600-merge.sh | 10 ++++++++++
 2 files changed, 11 insertions(+), 1 deletion(-)


base-commit: d63586cb314731c851f28e14fc8012988467e2da

Patch

diff --git a/builtin/merge.c b/builtin/merge.c
index daed2d4e1e2..03c405fa5df 100644
--- a/builtin/merge.c
+++ b/builtin/merge.c
@@ -701,7 +701,7 @@  static int try_merge_strategy(const char *strategy, struct commit_list *common,
 	if (repo_refresh_and_write_index(the_repository, REFRESH_QUIET,
 					 SKIP_IF_UNCHANGED, 0, NULL, NULL,
 					 NULL) < 0)
-		return error(_("Unable to write index."));
+		die(_("Unable to write index."));
 
 	if (!strcmp(strategy, "recursive") || !strcmp(strategy, "subtree") ||
 	    !strcmp(strategy, "ort")) {
diff --git a/t/t7600-merge.sh b/t/t7600-merge.sh
index e5ff073099a..ef54cff4faa 100755
--- a/t/t7600-merge.sh
+++ b/t/t7600-merge.sh
@@ -236,6 +236,16 @@  test_expect_success 'merge c1 with c2' '
 	verify_parents $c1 $c2
 '
 
+test_expect_success 'merge c1 with c2 when index.lock exists' '
+	test_when_finished rm .git/index.lock &&
+	git reset --hard c1 &&
+	>.git/index.lock &&
+	test_must_fail git merge c2 &&
+	test_path_is_missing .git/MERGE_HEAD &&
+	test_path_is_missing .git/MERGE_MODE &&
+	test_path_is_missing .git/MERGE_MSG
+'
+
 test_expect_success 'merge --squash c3 with c7' '
 	git reset --hard c3 &&
 	test_must_fail git merge --squash c7 &&