cache-tree: fix inverted object existence check in cache_tree_fully_valid
PATCH 1/1 cache-tree: fix inverted object existence check in cache_tree_fully_valid
cache_tree_fully_valid() is supposed to return 0 (not valid) when a tree object is missing from the object database. The condition currently returns 0 when odb_has_object() succeeds, which is the opposite of what is intended: the cache tree should be considered invalid when the object does not exist.
Add the missing negation so the function correctly invalidates cache tree nodes whose objects are absent.
--- cache-tree.c | 2 +- t/t0090-cache-tree.sh | 8 ++++++++ 2 files changed, 9 insertions(+), 1 deletion(-)
cache-tree.c +1 -1
@@ -238,7 +238,7 @@ int cache_tree_fully_valid(struct cache_tree *it)if (!it)return 0 ;if (it->entry_count <0 || -odb_has_object (the_repository->objects, &it->oid, + !odb_has_object (the_repository->objects, &it->oid,HAS_OBJECT_RECHECK_PACKED |HAS_OBJECT_FETCH_PROMISOR ))return 0 ;for (i =0 ; i < it->subtree_nr; i++) {
t/t0090-cache-tree.sh +8 -0
@@ -278,4 +278,12 @@ test_expect_success 'switching trees does not invalidate shared index' ' ) ' +test_expect_success 'cache-tree is used by write-tree when valid' ' + test_commit use-valid && + + # write-tree with a valid cache-tree should skip cache_tree_update + GIT_TRACE2_PERF="$(pwd)/trace.output" git write-tree && + ! grep region_enter.*cache_tree.*update trace.output +' + test_done
base-commit: 2855562ca6a9c6b0e7bc780b050c1e83c9fcfbd0
Signature · 1 line
2.52.0.ge17bebe515.stripe
Quoted 5 lines · David Lin
cache_tree_fully_valid()is supposed to return 0 (not valid) when a tree object is missing from the object database. The condition currently returns 0 whenodb_has_object()succeeds, which is the opposite of what is intended: the cache tree should be considered invalid when the object does not exist.
It looks like this negation occurred two refactors ago in 062b914c84 (treewide: convert users of repo_has_object_file() to has_object(), 2025-04-29) which had this diff:
cache-tree.c +3 -1
@@ -238,7 +238,9 @@ int cache_tree_fully_valid(struct cache_tree *it)int i;if (!it)return 0 ; -if (it->entry_count <0 || !repo_has_object_file (the_repository, &it->oid)) +if (it->entry_count <0 || +has_object (the_repository, &it->oid, +HAS_OBJECT_RECHECK_PACKED |HAS_OBJECT_FETCH_PROMISOR ))return 0 ;for (i =0 ; i < it->subtree_nr; i++) {if (!cache_tree_fully_valid (it->down[i]->cache_tree))
This was one hunk among many, so it is easy to miss that the ! was lost as the function was renamed and moved to another line.
Review
Quoted 5 lines
(quoted diff) +1
-1if (it->entry_count < 0 || - odb_has_object(the_repository->objects, &it->oid, + !odb_has_object(the_repository->objects, &it->oid, HAS_OBJECT_RECHECK_PACKED | HAS_OBJECT_FETCH_PROMISOR)) return 0;
This fix (respecting the has_object to odb_has_object refactor) is the correct one.
Review
Quoted 6 lines
(quoted diff) +6
-0+test_expect_success 'cache-tree is used by write-tree when valid' ' + test_commit use-valid && + + # write-tree with a valid cache-tree should skip cache_tree_update + GIT_TRACE2_PERF="$(pwd)/trace.output" git write-tree && + ! grep region_enter.*cache_tree.*update trace.output
nit: I think this would be better as "test_grep ! ..."
Thanks,
-Stolee
Quoted 27 lines · Derrick Stolee
Quoted 5 lines · David Lin
cache_tree_fully_valid()is supposed to return 0 (not valid) when a tree object is missing from the object database. The condition currently returns 0 whenodb_has_object()succeeds, which is the opposite of what is intended: the cache tree should be considered invalid when the object does not exist.It looks like this negation occurred two refactors ago in
062b914c84(treewide: convert users ofrepo_has_object_file()tohas_object(), 2025-04-29) which had this diff:cache-tree.c +3
-1@@ -238,7 +238,9 @@ int cache_tree_fully_valid(struct cache_tree *it)int i;if (!it)return 0 ; -if (it->entry_count <0 || !repo_has_object_file (the_repository, &it->oid)) +if (it->entry_count <0 || +has_object (the_repository, &it->oid, +HAS_OBJECT_RECHECK_PACKED |HAS_OBJECT_FETCH_PROMISOR ))return 0 ;for (i =0 ; i < it->subtree_nr; i++) {if (!cache_tree_fully_valid (it->down[i]->cache_tree))
Quoted 2 lines
This was one hunk among many, so it is easy to miss that the ! was lost as the function was renamed and moved to another line.
Thanks for archaeology. It deserves to be recorded in the proposed log message. David, would you mind updating the message and repost a v2 of the patch?
I wonder what the practical effect of this breakage was. The added test only checks what happens in the trace, but what was the effect
externally observable? We did not rebuild the cache-tree when we
should have, causing "write-tree" to record a set of tree objects that do not match what is in the index?
Quoted 20 lines
Review
Quoted 5 lines
(quoted diff) +1
-1if (it->entry_count < 0 || - odb_has_object(the_repository->objects, &it->oid, + !odb_has_object(the_repository->objects, &it->oid, HAS_OBJECT_RECHECK_PACKED | HAS_OBJECT_FETCH_PROMISOR)) return 0;This fix (respecting the
has_objecttoodb_has_objectrefactor) is the correct one.Review
Quoted 6 lines
(quoted diff) +6
-0+test_expect_success 'cache-tree is used by write-tree when valid' ' + test_commit use-valid && + + # write-tree with a valid cache-tree should skip cache_tree_update + GIT_TRACE2_PERF="$(pwd)/trace.output" git write-tree && + ! grep region_enter.*cache_tree.*update trace.outputnit: I think this would be better as "
test_grep! ..."Thanks,
-Stolee
PATCH v2 1/1 cache-tree: fix inverted object existence check in cache_tree_fully_valid
The negation in front of the object existence check in cache_tree_fully_valid() was lost in 062b914c84 (treewide: convert users of repo_has_object_file() to has_object(), 2025-04-29), turning !repo_has_object_file(...) into has_object(...) instead of !has_object(...).
This makes cache_tree_fully_valid() always report the cache tree as invalid when objects exist (the common case), forcing callers like write_index_as_tree() to call cache_tree_update() on every invocation. An odb_has_object() check inside update_one() avoids a full tree rebuild, but the unnecessary call still pays the cost of opening an ODB transaction and, in partial clones, a promisor remote check.
Restore the missing negation and add a test that verifies write-tree takes the cache-tree shortcut when the cache tree is valid.
---
Notes:
Changes since v1:
- Use test_grep instead of bare grep (Stolee)
cache-tree.c | 2 +- t/t0090-cache-tree.sh | 8 ++++++++ 2 files changed, 9 insertions(+), 1 deletion(-)
cache-tree.c +1 -1
@@ -238,7 +238,7 @@ int cache_tree_fully_valid(struct cache_tree *it)if (!it)return 0 ;if (it->entry_count <0 || -odb_has_object (the_repository->objects, &it->oid, + !odb_has_object (the_repository->objects, &it->oid,HAS_OBJECT_RECHECK_PACKED |HAS_OBJECT_FETCH_PROMISOR ))return 0 ;for (i =0 ; i < it->subtree_nr; i++) {
t/t0090-cache-tree.sh +8 -0
@@ -278,4 +278,12 @@ test_expect_success 'switching trees does not invalidate shared index' ' ) ' +test_expect_success 'cache-tree is used by write-tree when valid' ' + test_commit use-valid && + + # write-tree with a valid cache-tree should skip cache_tree_update + GIT_TRACE2_PERF="$(pwd)/trace.output" git write-tree && + test_grep ! region_enter.*cache_tree.*update trace.output +' + test_done
base-commit: 2855562ca6a9c6b0e7bc780b050c1e83c9fcfbd0
Signature · 1 line
2.52.0.ge17bebe515.stripe
Quoted 14 lines · David Lin
The negation in front of the object existence check in
cache_tree_fully_valid()was lost in062b914c84(treewide: convert users ofrepo_has_object_file()tohas_object(), 2025-04-29), turning!repo_has_object_file(...)intohas_object(...)instead of!has_object(...).This makes
cache_tree_fully_valid()always report the cache tree as invalid when objects exist (the common case), forcing callers likewrite_index_as_tree()to callcache_tree_update()on every invocation. Anodb_has_object()check insideupdate_one()avoids a full tree rebuild, but the unnecessary call still pays the cost of opening an ODB transaction and, in partial clones, a promisor remote check.
Quoted 2 lines
Restore the missing negation and add a test that verifies write-tree takes the cache-tree shortcut when the cache tree is valid.
Oh, indeed, thanks for the fix.
I also checked whether there's any other such case in the commit in question, but didn't spot any.
cache-tree.c int cache_tree_fully_valid(struct cache_tree *it) · line 244
Quoted 13 lines
cache-tree.c +1
-1@@ -238,7 +238,7 @@ int cache_tree_fully_valid(struct cache_tree *it)if (!it)return 0 ;if (it->entry_count <0 || -odb_has_object (the_repository->objects, &it->oid, + !odb_has_object (the_repository->objects, &it->oid,HAS_OBJECT_RECHECK_PACKED |HAS_OBJECT_FETCH_PROMISOR ))return 0 ;for (i =0 ; i < it->subtree_nr; i++) {
Yup, this looks obviously good to me.
Patrick