Fail closed on two corrupt node shapes the readers accepted by timsehn · Pull Request #2007 · dolthub/doltlite

timsehn · 2026-08-06T17:04:50Z

Last two of the three prolly-core items from the deep review (after #2002). Both are hostile-input hardening in the read path — the writer produces neither shape, and every other structural check passes them.

1. An internal node pointing at an empty leaf

descendToExtremeLeaf set idx = pChild->node.nItems - 1, which is -1 when walking right, and prollyCursorNext/Prev then marked the cursor valid over a node whose key array is a null pointer. The next key read dereferences it.

prollyCursorFirst and Last already refused this through finalizeExtremeCursor, so only the stepping paths were exposed. Measured on a crafted two-child tree (one populated leaf, one empty):

	before	after
`prollyCursorNext` past the populated leaf	`SQLITE_OK`, cursor marked valid	`SQLITE_CORRUPT`
`prollyCursorLast`	`SQLITE_OK`, silently EOF	`SQLITE_CORRUPT`

The check sits on the descent, so an empty root leaf — which is just an empty table, not corruption — still reads as EOF. prolly_cursor_empty_leaf_root pins that and still passes.

2. Offset arrays not starting at zero

Readers reach key data two ways: through the offset array (prollyNodeKey) and by striding from pKeyData for fixed-width keys (prollyNodeIntKey, i*8). A node with aKeyOff = [8, 16] satisfies the bounds check, monotonicity, the INTKEY stride check (16-8 == 8) and the total-size check — and the two readers then disagree about which bytes are the key. Silently wrong data rather than a rejected node.

Both arrays must now start at 0.

Testing

Two cases in test/doltlite_regression_test_c.c, beside the existing prolly_cursor_corrupt_node and prolly_node_corruption cases and following their hand-built-node style.

Fail-before/pass-after, same test file both runs:

prolly_node_first_offset_validation — 2/4 → 4/4
prolly_cursor_empty_leaf_under_internal — 8/10 → 10/10

Each includes a guard that the fix is not just rejecting everything: a well-formed node with zero-start offsets still parses, and First on the crafted tree still returns a valid cursor.

Validation

test/doltlite_regression_test_c.sh — 310,193 tests, 0 failures
test/run_c_tests.sh — 26/26 gated c-tests
test/run_doltlite_tests.sh — 99/99 suites
testfixture storage bucket — 45,265 passing, 0 unexpected failures (the 2 bigfile terminations are environmental and pre-existing, confirmed against a control build during Bound the chunk store WAL offset before trusting it #1987)
testfixture core-sql bucket — the only failures are 7 shell/loadext/zipfile files, and I diffed the exact failure sets against a control build without these changes: identical, 53 each

A note on that last point, since it nearly fooled me: my first bucket runs reported ~2300 failures with "no such table" everywhere. That was a stale testfixture — I had rebuilt doltlite and the library but not it. Rebuilt, affinity2 went from 5 failures to 26/26 clean. Worth remembering that this binary needs an explicit rebuild.

🤖 Generated with Claude Code

Both are hostile-input hardening in the read path: the writer produces neither shape, and every other structural check passes them. An internal node pointing at an empty leaf. descendToExtremeLeaf set idx to nItems-1, which is -1 when walking right, and prollyCursorNext/Prev then marked the cursor valid over a node whose key array is a null pointer -- the next key read dereferences it. prollyCursorFirst and Last already refused it through finalizeExtremeCursor, so only the stepping paths were exposed. The check goes on the descent, so an empty root leaf, which is just an empty table, still reads as EOF. Offset arrays not starting at zero. Readers reach key data two ways, through the offset array and by striding from pKeyData for fixed-width keys, as prollyNodeIntKey does with i*8. A first offset of 8 with a stride of 8 satisfies the bounds, monotonicity, width and total-size checks, and the two readers then disagree about which bytes are the key -- silently wrong data rather than a rejected node. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

itoqa · 2026-08-06T17:32:33Z

Commit: dec0a96: 14 test cases ran, 14 passed ✅.

Summary

Coverage focuses on safe handling of malformed and empty stored data, including traversal in both directions, cache loading, scanning, cursor cleanup, and validation of valid data alongside corruption cases. The exercised behavior is primarily defensive edge-case and adversarial data handling, with normal empty-data and valid-record behavior also confirmed.

Safe to merge — all exercised behaviors passed, including corruption detection, normal end-of-data handling, valid-record access, and safe cursor cleanup, with no PR-attributable regressions or unresolved failures. Any unrelated observations do not present a merge blocker.

Tests run by Ito

Result	Severity	Type	Description
✅	—	Corruption	Loading a malformed data node returned a corruption error and did not add the node to the cache. All 4 native checks passed.
✅	—	Corruption	The malformed-record scan stopped with a corruption error instead of completing as if the data were valid. All 11 native assertions passed, and no changes were emitted from the bad data.
✅	—	Corruption	Moving through malformed data in either direction returned an error instead of pretending the scan was finished. The native regression case completed all 10 checks successfully.
✅	—	Cursor	The cursor reaches the first item, then reports database corruption when moving into the empty child instead of treating it as a valid row or normal end of data.
✅	—	Cursor	An empty table opens normally, and moving to the first or last position returns EOF without an error.
✅	—	Cursor	Moving from a populated leaf into an empty child returns a corruption error, and the cursor is closed safely without exposing a row.
✅	—	Harness	The native regression case passed all 10 assertions. Moving forward from the populated leaf and checking the rightmost path both handled the empty child as corruption without crashing.
✅	—	Harness	The empty-root control stayed a valid empty table, and the empty-child case returned the expected corruption result. Both cases passed independently in sequence.
✅	—	Harness	Both named regression cases were found and completed successfully through the shell wrapper and the direct test binary. All four invocations exited successfully, so the new cases are registered and discoverable.
✅	—	Harness	Two complete sessions passed every malformed-offset, empty-root, and empty-child check. The test finished cleanly with no crash or cleanup failure.
✅	—	Offset	Malformed node data with a nonzero first value offset is rejected as corrupt, while a valid zero-start node still parses and exposes one item.
✅	—	Offset	A valid node with both offset lists starting at zero was accepted and exposed exactly one item with usable integer-key data.
✅	—	Offset	A malformed data node with a shifted key offset is rejected before it can expose inconsistent key data.
✅	—	Offset	An empty non-root node is rejected with a corruption error before any parsed data is exposed.

Tip

Reply with @itoqa to send us feedback on this test run.

The previous run's macOS and Windows builds were cancelled after several hours of queueing rather than failing, and re-running the failed jobs did not requeue them.

itoqa · 2026-08-06T23:50:11Z

Ito Diff Report — dec0a96 → 11cf0d7: 26 test cases ran, 25 passing ✅, 1 additional finding ⚠️.

Diff Summary

Coverage focuses on safe database reading across normal and hostile data conditions: valid records and scans, empty tables, tree-boundary movement, malformed serialized data, sparse reads, and cleanup after corruption. Overall behavior is healthy, with corruption generally surfaced safely while valid data remains readable and ordinary empty-table end-of-file behavior is preserved.

Safe to merge — the only failure is a medium-severity, pre-existing cursor-state safety issue explicitly marked unrelated to this PR, with no regressions or PR-attributable failures. It is appropriate as a flag for later rather than a merge blocker.

Tests run by Ito

Result	State	Severity	Type	Description
✅	Passing	—	Corruption	Malformed offset data is rejected with SQLITE_CORRUPT, while a valid node with zero-start offsets still loads successfully.
✅	Passing	—	Corruption	A SQL scan stops with a corruption error when it reaches malformed data instead of finishing successfully or returning a partial result as complete.
✅	Passing	—	Corruption	Moving across the malformed child reports a corruption error in both directions instead of pretending the scan reached the end.
✅	Passing	—	Corruption	Malformed data from a sparse read is rejected, and rejected entries are not kept for later use. The available parser control passed all four checks; the requested named case was not registered in the native test runner.
✅	Passing	—	Cursor	Moving past the last item in a populated leaf returns a corruption error instead of treating the empty child as a valid row or ordinary end of file.
✅	Passing	—	Cursor	An empty root opens normally. Moving to the first or last position returns end of file without exposing a row.
✅	Passing	—	Cursor	Reverse traversal into an empty child stops with a corruption error instead of treating the invalid child as a usable cursor.
✅	Passing	—	Cursor	A large populated tree was scanned in order, and the cursor reached ordinary EOF only after the final item.
✅	Passing	—	Cursor	A blob-key lookup crossed an internal boundary, found the requested key in the right leaf, and kept the cursor valid.
✅	Passing	—	Harness	The regression case reached the populated item, then correctly reported corruption when traversal entered an empty child from either direction. All 10 assertions passed.
✅	Passing	—	Harness	The empty table reaches normal end-of-file, while an empty child inside a larger tree is rejected as corruption. Both neighboring checks ran independently and passed.
✅	Passing	—	Harness	Both requested checks were found by name and ran successfully. The empty-child check passed all 10 assertions, and the first-offset check passed all 4 assertions.
✅	Passing	—	Harness	Two repeated native sessions checked malformed offsets, valid zero-start data, an empty root, and an empty child. All assertions passed, and both sessions exited cleanly without a crash or cleanup failure.
✅	Passing	—	Harness	The native regression case rejects malformed key and value offsets and still accepts the valid zero-start control. All four checks passed.
✅	Passing	—	Harness	The malformed offset checks passed all four assertions and the process exited cleanly without a crash or allocator error.
✅	Passing	—	Harness	Both named regression cases were found and ran their intended checks. The empty-child cursor case passed 10 checks, and the first-offset case passed 4 checks.
✅	Passing	—	Harness	The malformed and valid fixtures behaved as expected in two complete native regression sessions. Both sessions passed all 310193 assertions and ended without a crash or allocator diagnostic.
✅	Passing	—	Offset	The native parser rejected a malformed node whose first value offset was 1, instead of reading shifted value data or accepting the node.
✅	Passing	—	Offset	A valid one-item integer record was accepted and remained available for reading.
✅	Passing	—	Offset	A malformed integer-key node is rejected before it can expose inconsistent key data, while valid zero-start data still works.
✅	Passing	—	Offset	An empty node below the root is rejected as corrupt before any reader can use it.
✅	Passing	—	Offset	The valid zero-start node was accepted and exposed its single item, while the malformed offset checks also passed.
✅	Passing	—	Offset	A malformed node with a value offset starting at one was rejected as corruption, while the valid zero-offset control still parsed successfully.
✅	Passing	—	Offset	Malformed node data is rejected with a corruption error instead of being read.
✅	Passing	—	Offset	The parser rejected a malformed node whose first key offset started at 8, while the valid zero-start control still parsed and exposed one item.
⚠️	Additional Finding		Cursor	After the cursor reaches the final row, stepping into an empty child returns SQLITE_CORRUPT but does not make the cursor unusable. Reading the next key or value can therefore operate on a child with no rows.

Additional Findings Details

These findings are unrelated to the current changes but were observed during testing.

🟡 Corrupt cursor stays marked as usable

Severity: Medium
Description: After the cursor reaches the final row, stepping into an empty child returns SQLITE_CORRUPT but does not make the cursor unusable. Reading the next key or value can therefore operate on a child with no rows.
Impact: When a database contains a malformed child node, moving through its rows can leave the cursor looking valid. A later read may crash or access an invalid row instead of safely reporting the corruption.
Steps to Reproduce:
1. Build the native storage-engine regression runner with the prolly implementation enabled.
2. Create an internal tree with one populated leaf followed by an empty child, and position the cursor on the populated leaf.
3. Call the forward step that crosses into the empty child and confirm it returns SQLITE_CORRUPT.
4. Call a key or value accessor after that error and verify that the cursor is not still reported as valid and no row is exposed.
Stub / mock content: No stubs, mocks, or bypasses were applied for this test in the recorded run.
Code Analysis: src/prolly_cursor.c:149-167 implements descendToExtremeLeaf. When the descended child has zero items, line 163 returns SQLITE_CORRUPT immediately. That function does not change cur->eState. In prollyCursorNext at src/prolly_cursor.c:270-307, the boundary path increments the parent child index, assigns cur->iLevel, calls descendToExtremeLeaf at line 299, and returns the error at line 300 before the success-path assignment cur->eState = PROLLY_CURSOR_VALID at line 301. Because the cursor was VALID before the call, it remains VALID on this error path. The accessors at lines 414-438 assert that eState is VALID and then use the current leaf and index; prollyCursorKey, prollyCursorIntKey, and prollyCursorValue can consequently read the failed empty-child state. The targeted fix is to set eState to PROLLY_CURSOR_INVALID (and release or otherwise clear the failed descent state as required by the cursor ownership rules) before returning SQLITE_CORRUPT from the failed stepping path. The separate level-zero empty-root path remains ordinary EOF and must not be changed.

Evidence Package

text/CURSOR-3/CURSOR-3-corrupt-cursor-evidence.txt — 3 KB

Tip

Reply with @itoqa to send us feedback on this test run.

github-actions · 2026-08-06T23:58:32Z

DoltLite source coverage

Metric	Covered	Total	Coverage
Lines	43148	51121	84.40%
Branches	18948	28286	66.99%
Functions	2101	2300	91.35%

Merged 162 pooled raw profiles from the distributed Linux correctness jobs.

Per-file coverage (91 files)

File	Lines	Branches	Functions
`src/btree_orig_api.c`	85.87%	76.19%	87.18%
`src/chunk_file.c`	100.00%	100.00%	100.00%
`src/chunk_index.c`	88.76%	76.47%	100.00%
`src/chunk_refs.c`	99.35%	85.48%	100.00%
`src/chunk_staging.c`	87.50%	77.05%	94.12%
`src/chunk_store.c`	91.01%	75.00%	100.00%
`src/chunk_store_commit.c`	86.32%	66.96%	100.00%
`src/chunk_store_lock.c`	94.25%	77.47%	100.00%
`src/chunk_store_refs_api.c`	91.52%	77.40%	100.00%
`src/chunk_wal.c`	94.24%	75.00%	92.86%
`src/doltlite.c`	100.00%	100.00%	100.00%
`src/doltlite_add.c`	85.19%	67.71%	100.00%
`src/doltlite_ancestor.c`	88.68%	66.11%	100.00%
`src/doltlite_at.c`	77.98%	59.93%	94.12%
`src/doltlite_blame.c`	80.70%	58.94%	96.43%
`src/doltlite_branch.c`	86.78%	80.20%	100.00%
`src/doltlite_branches.c`	93.51%	77.27%	92.86%
`src/doltlite_checkout.c`	76.38%	59.81%	100.00%
`src/doltlite_cherry_pick.c`	70.53%	55.08%	71.43%
`src/doltlite_chunk_walk.c`	88.60%	64.19%	100.00%
`src/doltlite_cmd.c`	69.23%	60.83%	93.33%
`src/doltlite_commit.c`	93.22%	73.08%	100.00%
`src/doltlite_commit_ancestors.c`	90.65%	68.89%	92.31%
`src/doltlite_commit_cmd.c`	75.72%	69.42%	100.00%
`src/doltlite_config.c`	77.54%	71.43%	100.00%
`src/doltlite_conflicts.c`	81.83%	58.92%	93.44%
`src/doltlite_constraint_violations.c`	81.56%	55.91%	92.86%
`src/doltlite_core.c`	91.69%	72.22%	100.00%
`src/doltlite_creds.c`	84.01%	54.75%	89.74%
`src/doltlite_dbpage.c`	92.68%	73.21%	91.67%
`src/doltlite_diff.c`	87.88%	69.04%	96.43%
`src/doltlite_diff_stat.c`	94.15%	75.38%	95.56%
`src/doltlite_diff_table.c`	93.78%	68.45%	97.14%
`src/doltlite_gc.c`	77.11%	56.01%	96.30%
`src/doltlite_hashof.c`	75.64%	62.86%	94.44%
`src/doltlite_history.c`	85.61%	70.55%	100.00%
`src/doltlite_http_remote.c`	82.22%	57.17%	92.31%
`src/doltlite_ignore.c`	90.06%	66.67%	100.00%
`src/doltlite_log.c`	94.38%	70.16%	92.31%
`src/doltlite_merge.c`	96.94%	74.49%	100.00%
`src/doltlite_merge_cmd.c`	88.89%	73.50%	100.00%
`src/doltlite_merge_constraints.c`	88.25%	67.65%	100.00%
`src/doltlite_merge_constraints_check.c`	92.68%	65.76%	100.00%
`src/doltlite_merge_constraints_fk.c`	77.09%	55.86%	85.71%
`src/doltlite_merge_constraints_unique.c`	86.50%	64.37%	100.00%
`src/doltlite_merge_pass1.c`	91.02%	74.87%	100.00%
`src/doltlite_merge_pass2.c`	62.90%	57.69%	100.00%
`src/doltlite_merge_rows.c`	88.25%	69.37%	100.00%
`src/doltlite_merge_schema.c`	84.75%	63.03%	96.55%
`src/doltlite_merge_status.c`	91.98%	73.53%	92.31%
`src/doltlite_patch.c`	94.92%	74.25%	97.83%
`src/doltlite_rebase.c`	85.57%	59.77%	100.00%
`src/doltlite_record.c`	75.90%	57.54%	92.86%
`src/doltlite_ref.c`	96.97%	77.27%	100.00%
`src/doltlite_remote.c`	81.98%	63.65%	95.92%
`src/doltlite_remote_sql.c`	63.89%	55.98%	92.86%
`src/doltlite_remotesrv.c`	75.94%	64.97%	90.62%
`src/doltlite_reset.c`	86.10%	71.53%	100.00%
`src/doltlite_revert.c`	79.29%	68.18%	100.00%
`src/doltlite_schema_diff.c`	92.38%	69.46%	96.88%
`src/doltlite_schemas.c`	68.97%	44.87%	90.91%
`src/doltlite_status.c`	92.46%	71.84%	97.22%
`src/doltlite_tag.c`	80.07%	56.62%	93.33%
`src/doltlite_tls.c`	85.04%	62.73%	92.31%
`src/doltlite_verify_constraints.c`	74.11%	58.46%	100.00%
`src/doltlite_workspace.c`	89.46%	65.50%	100.00%
`src/pager_shim.c`	58.00%	64.73%	32.37%
`src/prolly_btree.c`	81.21%	63.06%	86.40%
`src/prolly_btree_catalog.c`	80.81%	66.17%	96.67%
`src/prolly_btree_cursor.c`	86.32%	62.69%	97.92%
`src/prolly_btree_cursor_count.c`	78.78%	55.70%	100.00%
`src/prolly_btree_cursor_payload.c`	78.81%	60.38%	94.12%
`src/prolly_btree_cursor_seek.c`	73.88%	61.69%	82.35%
`src/prolly_btree_mutation.c`	85.54%	65.05%	97.92%
`src/prolly_btree_orig.c`	83.44%	35.71%	87.67%
`src/prolly_btree_state.c`	93.89%	67.48%	100.00%
`src/prolly_btree_txn.c`	82.70%	69.84%	98.04%
`src/prolly_cache.c`	93.24%	69.70%	100.00%
`src/prolly_check.c`	60.36%	62.96%	100.00%
`src/prolly_chunker.c`	94.04%	78.38%	100.00%
`src/prolly_cursor.c`	89.65%	81.25%	100.00%
`src/prolly_diff.c`	52.48%	38.62%	60.87%
`src/prolly_hash.c`	93.65%	80.00%	100.00%
`src/prolly_hashset.c`	90.48%	80.56%	100.00%
`src/prolly_mutate.c`	87.07%	78.08%	100.00%
`src/prolly_mutmap.c`	92.95%	80.04%	100.00%
`src/prolly_node.c`	89.31%	74.24%	100.00%
`src/prolly_three_way_diff.c`	95.58%	85.90%	100.00%
`src/prolly_three_way_merge.c`	80.78%	66.67%	91.67%
`src/prolly_xxhash.c`	100.00%	100.00%	100.00%
`src/sortkey.c`	93.06%	80.14%	100.00%

Download HTML and LCOV artifacts from this workflow run.

github-actions · 2026-08-07T00:03:24Z

DoltLite performance vs PR base

Baseline: 69832551d03778e273ae168d1577d223d24e628e
Candidate: 1b9a0300de9afc67e98bebdad473a16e54ae826c
Overall ratio: 1.005x
Gate result: PASS
Gates: individual > 1.25x with more than 5.00ms regression; section, suite, or overall > 1.15x with the same minimum delta
vc individual gate: > 1.50x with more than 25.00ms regression
Confirmed failed gates: none
Automatic retries: textpk cleared after 2 attempts; no gate failed every time

Suite	Workloads	Baseline total	Candidate total	Ratio	Result
blobpk	69	11.88s	11.90s	1.001x	PASS
compositepk	69	12.20s	12.20s	0.999x	PASS
int	69	10.94s	10.95s	1.000x	PASS
textpk	69	10.35s	10.58s	1.023x	PASS
vc	13	911.71ms	918.81ms	1.008x	PASS

blobpk details

Section	Test	Baseline	Candidate	Delta	Ratio	Result
mem_reads	`oltp_point_select`	36.78ms	36.41ms	-365us	0.990x	PASS
mem_reads	`oltp_range_select`	13.86ms	13.91ms	+47us	1.003x	PASS
mem_reads	`oltp_sum_range`	13.88ms	13.96ms	+72us	1.005x	PASS
mem_reads	`oltp_order_range`	3.15ms	3.10ms	-42us	0.987x	PASS
mem_reads	`oltp_distinct_range`	4.24ms	4.21ms	-31us	0.993x	PASS
mem_reads	`oltp_index_scan`	5.93ms	6.01ms	+74us	1.012x	PASS
mem_reads	`select_random_points`	20.44ms	20.61ms	+167us	1.008x	PASS
mem_reads	`select_random_ranges`	5.01ms	5.12ms	+108us	1.022x	PASS
mem_reads	`covering_index_scan`	4.27ms	4.32ms	+49us	1.011x	PASS
mem_reads	`groupby_scan`	34.00ms	33.82ms	-173us	0.995x	PASS
mem_reads	`index_join`	8.65ms	8.79ms	+140us	1.016x	PASS
mem_reads	`index_join_scan`	5.46ms	5.48ms	+23us	1.004x	PASS
mem_reads	`types_table_scan`	1.26s	1.26s	-4.32ms	0.997x	PASS
mem_reads	`table_scan`	1.35s	1.36s	+11.02ms	1.008x	PASS
mem_reads	`oltp_read_only`	133.82ms	135.56ms	+1.74ms	1.013x	PASS
mem_writes	`oltp_bulk_insert`	350.04ms	347.30ms	-2.74ms	0.992x	PASS
mem_writes	`oltp_insert`	38.71ms	38.76ms	+50us	1.001x	PASS
mem_writes	`oltp_update_index`	126.21ms	126.15ms	-60us	1.000x	PASS
mem_writes	`oltp_update_non_index`	82.91ms	82.53ms	-377us	0.995x	PASS
mem_writes	`oltp_delete_insert`	100.74ms	100.32ms	-414us	0.996x	PASS
mem_writes	`oltp_write_only`	59.64ms	59.67ms	+31us	1.001x	PASS
mem_writes	`types_delete_insert`	52.54ms	52.67ms	+133us	1.003x	PASS
mem_writes	`oltp_read_write`	136.82ms	136.79ms	-34us	1.000x	PASS
file_reads	`oltp_point_select`	60.70ms	61.06ms	+358us	1.006x	PASS
file_reads	`oltp_range_select`	16.66ms	16.59ms	-76us	0.995x	PASS
file_reads	`oltp_sum_range`	16.60ms	16.51ms	-91us	0.995x	PASS
file_reads	`oltp_order_range`	3.46ms	3.44ms	-26us	0.992x	PASS
file_reads	`oltp_distinct_range`	4.54ms	4.53ms	-13us	0.997x	PASS
file_reads	`oltp_index_scan`	8.63ms	8.78ms	+150us	1.017x	PASS
file_reads	`select_random_points`	24.01ms	24.80ms	+791us	1.033x	PASS
file_reads	`select_random_ranges`	7.64ms	7.71ms	+73us	1.010x	PASS
file_reads	`covering_index_scan`	6.92ms	6.98ms	+62us	1.009x	PASS
file_reads	`groupby_scan`	34.67ms	34.46ms	-206us	0.994x	PASS
file_reads	`index_join`	10.93ms	11.11ms	+186us	1.017x	PASS
file_reads	`index_join_scan`	5.87ms	5.84ms	-33us	0.994x	PASS
file_reads	`types_table_scan`	1.26s	1.25s	-6.22ms	0.995x	PASS
file_reads	`table_scan`	1.35s	1.35s	+1.75ms	1.001x	PASS
file_reads	`oltp_read_only`	171.12ms	172.50ms	+1.38ms	1.008x	PASS
file_writes	`oltp_bulk_insert`	371.15ms	371.59ms	+445us	1.001x	PASS
file_writes	`oltp_insert`	50.31ms	50.86ms	+545us	1.011x	PASS
file_writes	`oltp_update_index`	159.89ms	159.75ms	-135us	0.999x	PASS
file_writes	`oltp_update_non_index`	104.81ms	104.72ms	-92us	0.999x	PASS
file_writes	`oltp_delete_insert`	126.02ms	126.70ms	+675us	1.005x	PASS
file_writes	`oltp_write_only`	81.58ms	80.51ms	-1.06ms	0.987x	PASS
file_writes	`types_delete_insert`	69.71ms	68.95ms	-757us	0.989x	PASS
file_writes	`oltp_read_write`	157.02ms	157.00ms	-16us	1.000x	PASS
ac_reads	`oltp_point_select`	60.79ms	60.77ms	-18us	1.000x	PASS
ac_reads	`oltp_range_select`	16.57ms	16.49ms	-74us	0.996x	PASS
ac_reads	`oltp_sum_range`	16.90ms	16.72ms	-182us	0.989x	PASS
ac_reads	`oltp_order_range`	3.57ms	3.44ms	-133us	0.963x	PASS
ac_reads	`oltp_distinct_range`	4.57ms	4.49ms	-75us	0.984x	PASS
ac_reads	`oltp_index_scan`	8.73ms	8.85ms	+122us	1.014x	PASS
ac_reads	`select_random_points`	24.22ms	24.05ms	-174us	0.993x	PASS
ac_reads	`select_random_ranges`	7.59ms	7.67ms	+72us	1.009x	PASS
ac_reads	`covering_index_scan`	7.01ms	6.98ms	-36us	0.995x	PASS
ac_reads	`groupby_scan`	34.83ms	34.26ms	-570us	0.984x	PASS
ac_reads	`index_join`	11.11ms	11.05ms	-60us	0.995x	PASS
ac_reads	`index_join_scan`	5.87ms	5.94ms	+68us	1.012x	PASS
ac_reads	`types_table_scan`	1.26s	1.26s	-1.86ms	0.999x	PASS
ac_reads	`table_scan`	1.35s	1.36s	+12.54ms	1.009x	PASS
ac_reads	`oltp_read_only`	170.54ms	169.89ms	-651us	0.996x	PASS
ac_writes	`oltp_bulk_insert_ac`	95.67ms	98.89ms	+3.22ms	1.034x	PASS
ac_writes	`oltp_insert_ac`	119.37ms	119.16ms	-208us	0.998x	PASS
ac_writes	`oltp_update_index_ac`	130.32ms	135.27ms	+4.95ms	1.038x	PASS
ac_writes	`oltp_update_non_index_ac`	111.00ms	110.87ms	-133us	0.999x	PASS
ac_writes	`oltp_delete_insert_ac`	125.17ms	123.39ms	-1.78ms	0.986x	PASS
ac_writes	`oltp_write_only_ac`	121.31ms	126.56ms	+5.25ms	1.043x	PASS
ac_writes	`types_delete_insert_ac`	115.73ms	111.12ms	-4.61ms	0.960x	PASS
ac_writes	`oltp_read_write_ac`	131.43ms	130.43ms	-995us	0.992x	PASS

compositepk details

Section	Test	Baseline	Candidate	Delta	Ratio	Result
mem_reads	`oltp_point_select`	40.37ms	40.51ms	+136us	1.003x	PASS
mem_reads	`oltp_range_select`	21.40ms	21.48ms	+74us	1.003x	PASS
mem_reads	`oltp_sum_range`	21.03ms	21.10ms	+67us	1.003x	PASS
mem_reads	`oltp_order_range`	3.85ms	3.90ms	+57us	1.015x	PASS
mem_reads	`oltp_distinct_range`	4.94ms	4.92ms	-21us	0.996x	PASS
mem_reads	`oltp_index_scan`	6.20ms	6.10ms	-101us	0.984x	PASS
mem_reads	`select_random_points`	32.10ms	31.93ms	-162us	0.995x	PASS
mem_reads	`select_random_ranges`	8.95ms	9.01ms	+62us	1.007x	PASS
mem_reads	`covering_index_scan`	4.29ms	4.16ms	-133us	0.969x	PASS
mem_reads	`groupby_scan`	39.16ms	39.10ms	-64us	0.998x	PASS
mem_reads	`index_join`	10.55ms	10.33ms	-220us	0.979x	PASS
mem_reads	`index_join_scan`	5.55ms	5.61ms	+61us	1.011x	PASS
mem_reads	`types_table_scan`	1.30s	1.29s	-9.27ms	0.993x	PASS
mem_reads	`table_scan`	1.39s	1.39s	+3.37ms	1.002x	PASS
mem_reads	`oltp_read_only`	170.23ms	169.15ms	-1.08ms	0.994x	PASS
mem_writes	`oltp_bulk_insert`	362.88ms	358.11ms	-4.78ms	0.987x	PASS
mem_writes	`oltp_insert`	36.78ms	36.80ms	+19us	1.001x	PASS
mem_writes	`oltp_update_index`	116.71ms	115.70ms	-1.01ms	0.991x	PASS
mem_writes	`oltp_update_non_index`	83.62ms	82.61ms	-1.01ms	0.988x	PASS
mem_writes	`oltp_delete_insert`	95.81ms	94.86ms	-950us	0.990x	PASS
mem_writes	`oltp_write_only`	57.57ms	58.08ms	+515us	1.009x	PASS
mem_writes	`types_delete_insert`	54.12ms	54.76ms	+637us	1.012x	PASS
mem_writes	`oltp_read_write`	154.84ms	154.82ms	-14us	1.000x	PASS
file_reads	`oltp_point_select`	64.72ms	65.16ms	+438us	1.007x	PASS
file_reads	`oltp_range_select`	23.98ms	24.21ms	+232us	1.010x	PASS
file_reads	`oltp_sum_range`	24.12ms	24.06ms	-57us	0.998x	PASS
file_reads	`oltp_order_range`	4.26ms	4.30ms	+40us	1.009x	PASS
file_reads	`oltp_distinct_range`	5.41ms	5.35ms	-63us	0.988x	PASS
file_reads	`oltp_index_scan`	9.20ms	9.11ms	-87us	0.991x	PASS
file_reads	`select_random_points`	36.12ms	36.77ms	+652us	1.018x	PASS
file_reads	`select_random_ranges`	11.74ms	11.89ms	+152us	1.013x	PASS
file_reads	`covering_index_scan`	7.15ms	7.05ms	-102us	0.986x	PASS
file_reads	`groupby_scan`	40.05ms	39.83ms	-218us	0.995x	PASS
file_reads	`index_join`	13.29ms	13.19ms	-104us	0.992x	PASS
file_reads	`index_join_scan`	6.16ms	6.17ms	+5us	1.001x	PASS
file_reads	`types_table_scan`	1.30s	1.30s	-3.49ms	0.997x	PASS
file_reads	`table_scan`	1.39s	1.39s	+1.95ms	1.001x	PASS
file_reads	`oltp_read_only`	210.41ms	210.15ms	-270us	0.999x	PASS
file_writes	`oltp_bulk_insert`	382.76ms	382.42ms	-342us	0.999x	PASS
file_writes	`oltp_insert`	46.34ms	46.26ms	-84us	0.998x	PASS
file_writes	`oltp_update_index`	144.98ms	145.39ms	+411us	1.003x	PASS
file_writes	`oltp_update_non_index`	105.77ms	106.18ms	+410us	1.004x	PASS
file_writes	`oltp_delete_insert`	120.77ms	120.48ms	-296us	0.998x	PASS
file_writes	`oltp_write_only`	78.55ms	77.67ms	-879us	0.989x	PASS
file_writes	`types_delete_insert`	69.32ms	68.99ms	-329us	0.995x	PASS
file_writes	`oltp_read_write`	178.52ms	177.62ms	-901us	0.995x	PASS
ac_reads	`oltp_point_select`	65.11ms	64.91ms	-200us	0.997x	PASS
ac_reads	`oltp_range_select`	24.10ms	24.28ms	+179us	1.007x	PASS
ac_reads	`oltp_sum_range`	24.00ms	24.12ms	+120us	1.005x	PASS
ac_reads	`oltp_order_range`	4.30ms	4.36ms	+59us	1.014x	PASS
ac_reads	`oltp_distinct_range`	5.33ms	5.32ms	-11us	0.998x	PASS
ac_reads	`oltp_index_scan`	8.90ms	8.91ms	+9us	1.001x	PASS
ac_reads	`select_random_points`	35.89ms	36.02ms	+123us	1.003x	PASS
ac_reads	`select_random_ranges`	11.86ms	11.67ms	-194us	0.984x	PASS
ac_reads	`covering_index_scan`	7.03ms	7.05ms	+24us	1.003x	PASS
ac_reads	`groupby_scan`	40.19ms	40.08ms	-107us	0.997x	PASS
ac_reads	`index_join`	12.90ms	12.98ms	+78us	1.006x	PASS
ac_reads	`index_join_scan`	6.14ms	6.03ms	-106us	0.983x	PASS
ac_reads	`types_table_scan`	1.29s	1.29s	+799us	1.001x	PASS
ac_reads	`table_scan`	1.38s	1.38s	+1.22ms	1.001x	PASS
ac_reads	`oltp_read_only`	207.13ms	208.48ms	+1.36ms	1.007x	PASS
ac_writes	`oltp_bulk_insert_ac`	81.45ms	80.82ms	-627us	0.992x	PASS
ac_writes	`oltp_insert_ac`	99.67ms	100.99ms	+1.32ms	1.013x	PASS
ac_writes	`oltp_update_index_ac`	111.12ms	110.85ms	-266us	0.998x	PASS
ac_writes	`oltp_update_non_index_ac`	92.88ms	91.25ms	-1.63ms	0.982x	PASS
ac_writes	`oltp_delete_insert_ac`	103.79ms	103.64ms	-151us	0.999x	PASS
ac_writes	`oltp_write_only_ac`	102.04ms	101.39ms	-648us	0.994x	PASS
ac_writes	`types_delete_insert_ac`	97.06ms	100.60ms	+3.54ms	1.037x	PASS
ac_writes	`oltp_read_write_ac`	109.45ms	112.91ms	+3.46ms	1.032x	PASS

int details

Section	Test	Baseline	Candidate	Delta	Ratio	Result
mem_reads	`oltp_point_select`	26.98ms	27.23ms	+248us	1.009x	PASS
mem_reads	`oltp_range_select`	11.73ms	11.84ms	+116us	1.010x	PASS
mem_reads	`oltp_sum_range`	11.24ms	11.22ms	-16us	0.999x	PASS
mem_reads	`oltp_order_range`	2.85ms	2.83ms	-22us	0.992x	PASS
mem_reads	`oltp_distinct_range`	3.90ms	3.84ms	-61us	0.984x	PASS
mem_reads	`oltp_index_scan`	4.70ms	4.71ms	+9us	1.002x	PASS
mem_reads	`select_random_points`	10.82ms	10.74ms	-85us	0.992x	PASS
mem_reads	`select_random_ranges`	3.91ms	3.95ms	+34us	1.009x	PASS
mem_reads	`covering_index_scan`	3.99ms	3.99ms	+1us	1.000x	PASS
mem_reads	`groupby_scan`	34.23ms	34.18ms	-47us	0.999x	PASS
mem_reads	`index_join`	7.55ms	7.55ms	-6us	0.999x	PASS
mem_reads	`index_join_scan`	4.58ms	4.56ms	-19us	0.996x	PASS
mem_reads	`types_table_scan`	1.26s	1.26s	-1.49ms	0.999x	PASS
mem_reads	`table_scan`	1.37s	1.37s	+1.33ms	1.001x	PASS
mem_reads	`oltp_read_only`	114.19ms	115.32ms	+1.13ms	1.010x	PASS
mem_writes	`oltp_bulk_insert`	239.81ms	241.43ms	+1.62ms	1.007x	PASS
mem_writes	`oltp_insert`	28.16ms	27.98ms	-182us	0.994x	PASS
mem_writes	`oltp_update_index`	104.14ms	104.22ms	+77us	1.001x	PASS
mem_writes	`oltp_update_non_index`	58.40ms	58.45ms	+57us	1.001x	PASS
mem_writes	`oltp_delete_insert`	77.67ms	77.36ms	-309us	0.996x	PASS
mem_writes	`oltp_write_only`	44.27ms	44.90ms	+626us	1.014x	PASS
mem_writes	`types_delete_insert`	39.54ms	39.21ms	-330us	0.992x	PASS
mem_writes	`oltp_read_write`	103.71ms	103.87ms	+155us	1.001x	PASS
file_reads	`oltp_point_select`	55.62ms	55.39ms	-230us	0.996x	PASS
file_reads	`oltp_range_select`	14.68ms	14.71ms	+33us	1.002x	PASS
file_reads	`oltp_sum_range`	14.29ms	14.30ms	+16us	1.001x	PASS
file_reads	`oltp_order_range`	3.21ms	3.19ms	-24us	0.993x	PASS
file_reads	`oltp_distinct_range`	4.14ms	4.25ms	+108us	1.026x	PASS
file_reads	`oltp_index_scan`	7.90ms	7.89ms	-8us	0.999x	PASS
file_reads	`select_random_points`	13.94ms	14.06ms	+130us	1.009x	PASS
file_reads	`select_random_ranges`	6.86ms	6.73ms	-127us	0.981x	PASS
file_reads	`covering_index_scan`	7.13ms	7.14ms	+14us	1.002x	PASS
file_reads	`groupby_scan`	34.83ms	34.57ms	-254us	0.993x	PASS
file_reads	`index_join`	9.47ms	9.61ms	+140us	1.015x	PASS
file_reads	`index_join_scan`	5.03ms	4.99ms	-40us	0.992x	PASS
file_reads	`types_table_scan`	1.26s	1.25s	-1.27ms	0.999x	PASS
file_reads	`table_scan`	1.37s	1.37s	-613us	1.000x	PASS
file_reads	`oltp_read_only`	154.59ms	156.52ms	+1.93ms	1.012x	PASS
file_writes	`oltp_bulk_insert`	257.92ms	256.68ms	-1.25ms	0.995x	PASS
file_writes	`oltp_insert`	35.42ms	35.31ms	-115us	0.997x	PASS
file_writes	`oltp_update_index`	127.13ms	126.35ms	-776us	0.994x	PASS
file_writes	`oltp_update_non_index`	80.58ms	80.13ms	-446us	0.994x	PASS
file_writes	`oltp_delete_insert`	95.79ms	95.84ms	+48us	1.001x	PASS
file_writes	`oltp_write_only`	64.53ms	64.25ms	-279us	0.996x	PASS
file_writes	`types_delete_insert`	52.12ms	52.44ms	+318us	1.006x	PASS
file_writes	`oltp_read_write`	123.30ms	123.10ms	-196us	0.998x	PASS
ac_reads	`oltp_point_select`	55.70ms	55.58ms	-112us	0.998x	PASS
ac_reads	`oltp_range_select`	14.69ms	14.73ms	+37us	1.003x	PASS
ac_reads	`oltp_sum_range`	14.46ms	14.19ms	-271us	0.981x	PASS
ac_reads	`oltp_order_range`	3.21ms	3.22ms	+5us	1.002x	PASS
ac_reads	`oltp_distinct_range`	4.22ms	4.20ms	-24us	0.994x	PASS
ac_reads	`oltp_index_scan`	7.90ms	7.87ms	-37us	0.995x	PASS
ac_reads	`select_random_points`	14.05ms	13.93ms	-119us	0.992x	PASS
ac_reads	`select_random_ranges`	6.76ms	6.85ms	+91us	1.013x	PASS
ac_reads	`covering_index_scan`	7.05ms	7.10ms	+45us	1.006x	PASS
ac_reads	`groupby_scan`	34.83ms	34.71ms	-120us	0.997x	PASS
ac_reads	`index_join`	9.50ms	9.53ms	+37us	1.004x	PASS
ac_reads	`index_join_scan`	5.00ms	4.98ms	-16us	0.997x	PASS
ac_reads	`types_table_scan`	1.25s	1.26s	+5.47ms	1.004x	PASS
ac_reads	`table_scan`	1.37s	1.37s	+2.06ms	1.002x	PASS
ac_reads	`oltp_read_only`	155.12ms	154.73ms	-387us	0.998x	PASS
ac_writes	`oltp_bulk_insert_ac`	58.58ms	58.57ms	-11us	1.000x	PASS
ac_writes	`oltp_insert_ac`	75.36ms	80.02ms	+4.66ms	1.062x	PASS
ac_writes	`oltp_update_index_ac`	97.25ms	91.52ms	-5.73ms	0.941x	PASS
ac_writes	`oltp_update_non_index_ac`	70.37ms	70.24ms	-125us	0.998x	PASS
ac_writes	`oltp_delete_insert_ac`	84.49ms	85.11ms	+618us	1.007x	PASS
ac_writes	`oltp_write_only_ac`	81.92ms	80.10ms	-1.82ms	0.978x	PASS
ac_writes	`types_delete_insert_ac`	74.64ms	71.68ms	-2.96ms	0.960x	PASS
ac_writes	`oltp_read_write_ac`	87.43ms	87.44ms	+6us	1.000x	PASS

textpk details

Section	Test	Baseline	Candidate	Delta	Ratio	Result
mem_reads	`oltp_point_select`	27.49ms	27.21ms	-281us	0.990x	PASS
mem_reads	`oltp_range_select`	10.74ms	10.80ms	+62us	1.006x	PASS
mem_reads	`oltp_sum_range`	10.64ms	10.57ms	-64us	0.994x	PASS
mem_reads	`oltp_order_range`	2.51ms	2.51ms	-2us	0.999x	PASS
mem_reads	`oltp_distinct_range`	3.29ms	3.27ms	-12us	0.996x	PASS
mem_reads	`oltp_index_scan`	4.67ms	4.59ms	-76us	0.984x	PASS
mem_reads	`select_random_points`	16.20ms	15.89ms	-311us	0.981x	PASS
mem_reads	`select_random_ranges`	4.18ms	4.14ms	-40us	0.990x	PASS
mem_reads	`covering_index_scan`	3.42ms	3.42ms	-1us	1.000x	PASS
mem_reads	`groupby_scan`	27.75ms	27.62ms	-129us	0.995x	PASS
mem_reads	`index_join`	6.75ms	6.98ms	+226us	1.033x	PASS
mem_reads	`index_join_scan`	4.55ms	4.51ms	-38us	0.992x	PASS
mem_reads	`types_table_scan`	972.93ms	972.42ms	-506us	0.999x	PASS
mem_reads	`table_scan`	1.08s	1.06s	-13.02ms	0.988x	PASS
mem_reads	`oltp_read_only`	101.28ms	100.91ms	-378us	0.996x	PASS
mem_writes	`oltp_bulk_insert`	263.38ms	260.03ms	-3.35ms	0.987x	PASS
mem_writes	`oltp_insert`	30.61ms	30.19ms	-427us	0.986x	PASS
mem_writes	`oltp_update_index`	106.00ms	105.71ms	-293us	0.997x	PASS
mem_writes	`oltp_update_non_index`	66.83ms	67.05ms	+223us	1.003x	PASS
mem_writes	`oltp_delete_insert`	80.38ms	80.36ms	-17us	1.000x	PASS
mem_writes	`oltp_write_only`	47.82ms	47.84ms	+20us	1.000x	PASS
mem_writes	`types_delete_insert`	40.75ms	40.62ms	-126us	0.997x	PASS
mem_writes	`oltp_read_write`	103.68ms	103.02ms	-664us	0.994x	PASS
file_reads	`oltp_point_select`	50.35ms	50.34ms	-14us	1.000x	PASS
file_reads	`oltp_range_select`	13.20ms	13.19ms	-15us	0.999x	PASS
file_reads	`oltp_sum_range`	13.10ms	12.87ms	-231us	0.982x	PASS
file_reads	`oltp_order_range`	2.82ms	2.81ms	-6us	0.998x	PASS
file_reads	`oltp_distinct_range`	3.55ms	3.58ms	+32us	1.009x	PASS
file_reads	`oltp_index_scan`	7.39ms	7.33ms	-61us	0.992x	PASS
file_reads	`select_random_points`	18.75ms	18.47ms	-285us	0.985x	PASS
file_reads	`select_random_ranges`	6.53ms	6.50ms	-30us	0.995x	PASS
file_reads	`covering_index_scan`	6.17ms	6.24ms	+61us	1.010x	PASS
file_reads	`groupby_scan`	28.14ms	28.09ms	-47us	0.998x	PASS
file_reads	`index_join`	9.07ms	9.05ms	-22us	0.998x	PASS
file_reads	`index_join_scan`	5.17ms	5.10ms	-72us	0.986x	PASS
file_reads	`types_table_scan`	970.23ms	974.56ms	+4.32ms	1.004x	PASS
file_reads	`table_scan`	1.08s	1.08s	-3.64ms	0.997x	PASS
file_reads	`oltp_read_only`	135.18ms	135.06ms	-121us	0.999x	PASS
file_writes	`oltp_bulk_insert`	343.17ms	349.39ms	+6.23ms	1.018x	PASS
file_writes	`oltp_insert`	58.49ms	58.27ms	-225us	0.996x	PASS
file_writes	`oltp_update_index`	202.84ms	233.38ms	+30.54ms	1.151x	PASS
file_writes	`oltp_update_non_index`	138.83ms	184.38ms	+45.55ms	1.328x	TRANSIENT
file_writes	`oltp_delete_insert`	168.14ms	164.37ms	-3.77ms	0.978x	PASS
file_writes	`oltp_write_only`	118.12ms	118.94ms	+825us	1.007x	PASS
file_writes	`types_delete_insert`	87.85ms	98.76ms	+10.91ms	1.124x	PASS
file_writes	`oltp_read_write`	171.70ms	180.85ms	+9.15ms	1.053x	PASS
ac_reads	`oltp_point_select`	51.07ms	51.12ms	+53us	1.001x	PASS
ac_reads	`oltp_range_select`	13.46ms	13.34ms	-122us	0.991x	PASS
ac_reads	`oltp_sum_range`	13.45ms	13.33ms	-123us	0.991x	PASS
ac_reads	`oltp_order_range`	2.83ms	2.81ms	-27us	0.990x	PASS
ac_reads	`oltp_distinct_range`	3.56ms	3.59ms	+34us	1.010x	PASS
ac_reads	`oltp_index_scan`	7.43ms	7.39ms	-35us	0.995x	PASS
ac_reads	`select_random_points`	19.04ms	18.87ms	-166us	0.991x	PASS
ac_reads	`select_random_ranges`	6.49ms	6.55ms	+56us	1.009x	PASS
ac_reads	`covering_index_scan`	6.20ms	6.19ms	-9us	0.999x	PASS
ac_reads	`groupby_scan`	28.48ms	28.27ms	-203us	0.993x	PASS
ac_reads	`index_join`	9.09ms	9.21ms	+120us	1.013x	PASS
ac_reads	`index_join_scan`	5.12ms	5.05ms	-73us	0.986x	PASS
ac_reads	`types_table_scan`	965.75ms	978.92ms	+13.18ms	1.014x	PASS
ac_reads	`table_scan`	1.07s	1.08s	+7.91ms	1.007x	PASS
ac_reads	`oltp_read_only`	134.46ms	135.17ms	+703us	1.005x	PASS
ac_writes	`oltp_bulk_insert_ac`	107.33ms	80.54ms	-26.79ms	0.750x	PASS
ac_writes	`oltp_insert_ac`	173.09ms	211.55ms	+38.45ms	1.222x	PASS
ac_writes	`oltp_update_index_ac`	193.01ms	107.68ms	-85.33ms	0.558x	PASS
ac_writes	`oltp_update_non_index_ac`	233.89ms	151.77ms	-82.12ms	0.649x	PASS
ac_writes	`oltp_delete_insert_ac`	129.89ms	412.35ms	+282.46ms	3.175x	TRANSIENT
ac_writes	`oltp_write_only_ac`	131.96ms	126.21ms	-5.75ms	0.956x	PASS
ac_writes	`types_delete_insert_ac`	93.08ms	201.91ms	+108.83ms	2.169x	TRANSIENT
ac_writes	`oltp_read_write_ac`	292.65ms	195.76ms	-96.90ms	0.669x	PASS

vc details

Section	Test	Baseline	Candidate	Delta	Ratio	Result
vc	`status_clean_many_tables`	80.58ms	80.59ms	+5us	1.000x	PASS
vc	`status_dirty_many_tables`	83.95ms	84.35ms	+401us	1.005x	PASS
vc	`diff_regular_working_one_table`	77.30ms	77.18ms	-118us	0.998x	PASS
vc	`diff_regular_working_many_tables`	88.89ms	89.75ms	+857us	1.010x	PASS
vc	`diff_stat_working_many_tables`	89.35ms	90.25ms	+897us	1.010x	PASS
vc	`diff_schema_working_many_tables`	90.16ms	90.55ms	+389us	1.004x	PASS
vc	`branch_list_many_branches`	21.78ms	21.73ms	-52us	0.998x	PASS
vc	`branch_create_delete`	24.30ms	24.46ms	+165us	1.007x	PASS
vc	`checkout_branch_clean`	54.61ms	54.98ms	+376us	1.007x	PASS
vc	`merge_data_no_conflicts`	28.40ms	28.62ms	+220us	1.008x	PASS
vc	`merge_schema_no_conflicts`	21.75ms	21.70ms	-50us	0.998x	PASS
vc	`merge_data_conflicts`	125.18ms	127.41ms	+2.23ms	1.018x	PASS
vc	`merge_data_conflicts_with_resolve`	125.47ms	127.24ms	+1.77ms	1.014x	PASS

All relative performance gates passed.

timsehn mentioned this pull request Aug 6, 2026

Forward incremental blob writes to the legacy btree #2008

Merged

Retrigger CI

11cf0d7

The previous run's macOS and Windows builds were cancelled after several hours of queueing rather than failing, and re-running the failed jobs did not requeue them.

timsehn mentioned this pull request Aug 6, 2026

Run the regression suite under assertions in CI, and fix what it caught #2010

Open

timsehn merged commit c44ac6e into master Aug 7, 2026
59 checks passed

timsehn mentioned this pull request Aug 7, 2026

Invalidate the cursor when a prolly descent hits an empty child #2011

Merged

Navigation Menu

Uh oh!

Fail closed on two corrupt node shapes the readers accepted - #2007

Fail closed on two corrupt node shapes the readers accepted#2007
timsehn merged 2 commits into
masterfrom
fix/prolly-node-cursor-corrupt-input

timsehn commented Aug 6, 2026

Uh oh!

itoqa Bot commented Aug 6, 2026

Uh oh!

itoqa Bot commented Aug 6, 2026

Uh oh!

github-actions Bot commented Aug 6, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 7, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

timsehn commented Aug 6, 2026

1. An internal node pointing at an empty leaf

2. Offset arrays not starting at zero

Testing

Validation

Uh oh!

itoqa Bot commented Aug 6, 2026

Uh oh!

itoqa Bot commented Aug 6, 2026

Uh oh!

github-actions Bot commented Aug 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

DoltLite source coverage

Uh oh!

github-actions Bot commented Aug 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

DoltLite performance vs PR base

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

github-actions Bot commented Aug 6, 2026 •

edited

Loading

github-actions Bot commented Aug 7, 2026 •

edited

Loading