fix(datacell): fix memory leak in FlattenDataCell query on exception by Copilot · Pull Request #1680 · antgroup/vsag

Copilot · 2026-03-12T02:33:42Z

When MemoryBlockIO::DirectReadImpl allocates a buffer across block boundaries (need_release = true), exceptions thrown by subsequent GetCodesById calls or ComputeDistsBatch4/ComputeDist skip the corresponding Release() calls, leaking memory.

Changes

query() batch-of-4 loop: Initialize all four code pointers to nullptr; wrap GetCodesById × 4 + ComputeDistsBatch4 in try/catch with a release_batch lambda that null-guards each pointer before releasing.
query() single-item loop: Wrap GetCodesById + ComputeDist in try/catch; release on both exception and normal paths with consistent release && codes guard.
ComputePairVectors(): Same pattern — initialize to nullptr, release_pair lambda, try/catch guarantees cleanup on exception.

// Before — leaks codes1..4 if ComputeDistsBatch4 (or any GetCodesById) throws
const auto* codes1 = this->GetCodesById(idx[i],     release1);
const auto* codes2 = this->GetCodesById(idx[i + 1], release2);
...
computer->ComputeDistsBatch4(codes1, codes2, codes3, codes4, ...);
if (release1) this->io_->Release(codes1);  // unreachable on exception

// After — guaranteed release via lambda + try/catch
auto release_batch = [&]() {
    if (release1 && codes1) this->io_->Release(codes1);
    ...
};
try {
    codes1 = this->GetCodesById(idx[i], release1);
    ...
    computer->ComputeDistsBatch4(...);
} catch (...) { release_batch(); throw; }
release_batch();

Original prompt

This section details on the original issue you should resolve

<issue_title>HGraph memory leak</issue_title>
<issue_description>Describe the bug
leak backtrace
vsag::SafeAllocator::Allocate(unsigned long) at ??:?
vsag::MemoryBlockIO::DirectReadImpl(unsigned long, unsigned long, bool&) const at ??:?
vsag::FlattenDataCell<vsag::FP32Quantizer<(vsag::MetricType)0>, vsag::MemoryBlockIO>::query(float*, vsag::Computer<vsag::FP32Quantizer<(vsag::MetricType)0> >*, unsigned int const*, unsigned int, vsag::QueryContext*) at ??:?
std::shared_ptr<vsag::DistanceHeap> vsag::BasicSearcher::search_impl<(vsag::InnerSearchMode)1>(std::shared_ptr<vsag::GraphInterface> const&, std::shared_ptr<vsag::FlattenInterface> const&, std::shared_ptr<vsag::VisitedList> const&, void const*, vsag::InnerSearchParam const&, std::shared_ptr<vsag::LabelTable> const&, vsag::QueryContext*) const at ??:?
vsag::BasicSearcher::Search(std::shared_ptr<vsag::GraphInterface> const&, std::shared_ptr<vsag::FlattenInterface> const&, std::shared_ptr<vsag::VisitedList> const&, void const*, vsag::InnerSearchParam const&, std::shared_ptr<vsag::LabelTable> const&, vsag::QueryContext*) const [clone .localalias] at xpointer.c:?
std::shared_ptr<vsag::DistanceHeap> vsag::HGraph::search_one_graph<(vsag::InnerSearchMode)1>(void const*, std::shared_ptr<vsag::GraphInterface> const&, std::shared_ptr<vsag::FlattenInterface> const&, vsag::InnerSearchParam&, std::shared_ptr<vsag::VisitedList> const&, vsag::QueryContext*) const at ??:?
vsag::HGraph::graph_add_one(void const*, int, unsigned int) at ??:?
vsag::HGraph::add_one_point(void const*, int, unsigned int) at ??:?
vsag::HGraph::Add(std::shared_ptr<vsag::Dataset> const&) at ??:?
vsag::IndexImpl<vsag::HGraph>::Add(std::shared_ptr<vsag::Dataset> const&) at ??:?
Analysis
Allocation and Deallocation Logic

Allocation Location (memory_block_io.cpp) memory_block_io.cpp

Release position (flatten_datacell.h) flatten_datacell.h Lines 354-391

Between GetCodesById and Release, ComputeDistsBatch4/ComputeDist is called. Once these functions throw an exception, Release will not be executed, and the allocated 1-4 buffers will leak.

To Reproduce
Codes to reproduce the behavior:
// paste codes here
Environment
Please run bash scripts/check_environment.sh and paste the output here:

OS: Linux

vsag version: v0.18.2

compiler version: e.g. GCC9

interface: cpp

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.
</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes HGraph memory leak #1634

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

…ception Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com>

Copilot

Pull request overview

Fixes an exception-path memory leak in FlattenDataCell queries by ensuring buffers returned from MemoryBlockIO::DirectReadImpl are always released even when downstream distance computations (or subsequent reads) throw.

Changes:

Add null-safe cleanup lambdas and try/catch guards around the batch-of-4 query loop to guarantee Release() runs on exceptions.
Add exception-safe cleanup to the single-item query loop with consistent release && codes guarding.
Wrap ComputePairVectors() body to ensure both code buffers are released on any exception.

Copilot · 2026-03-12T07:46:52Z

src/datacell/flatten_datacell.h

+        try {
+            codes1 = this->GetCodesById(idx[i], release1);
+            codes2 = this->GetCodesById(idx[i + 1], release2);
+            codes3 = this->GetCodesById(idx[i + 2], release3);
+            codes4 = this->GetCodesById(idx[i + 3], release4);
+            computer->ComputeDistsBatch4(codes1,
+                                         codes2,
+                                         codes3,
+                                         codes4,
+                                         result_dists[i],
+                                         result_dists[i + 1],
+                                         result_dists[i + 2],
+                                         result_dists[i + 3]);
+        } catch (...) {
+            release_batch();
+            throw;
        }
+        release_batch();
    }
    for (; i < id_count; ++i) {
        bool release = false;
-        const auto* codes = this->GetCodesById(idx[i], release);
-        computer->ComputeDist(codes, result_dists + i);
-        if (release) {
+        const uint8_t* codes = nullptr;
+        try {
+            codes = this->GetCodesById(idx[i], release);
+            computer->ComputeDist(codes, result_dists + i);
+        } catch (...) {
+            if (release && codes) {
+                this->io_->Release(codes);
+            }
+            throw;
+        }
+        if (release && codes) {
            this->io_->Release(codes);
        }


Missing regression test coverage for the new exception-safety paths. The leak fix is specifically about exceptions thrown between GetCodesById() and Release(), but current unit tests for FlattenDataCell appear to exercise only the non-throwing query path. Please add a Catch2 test that forces need_release=true (e.g., by using block_memory_io with a small block size / boundary-crossing reads) and then injects an exception from ComputeDist/ComputeDistsBatch4 (or an equivalent hook) to assert the allocated buffers are released (e.g., via a counting allocator or LSAN/ASAN expectation).

Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com>

inabao

LGTM

codecov · 2026-03-12T09:34:24Z

Codecov Report

❌ Patch coverage is 68.62745% with 16 lines in your changes missing coverage. Please review.

@@            Coverage Diff             @@
##             main    #1680      +/-   ##
==========================================
+ Coverage   90.66%   90.91%   +0.25%     
==========================================
  Files         329      329              
  Lines       19455    19474      +19     
==========================================
+ Hits        17639    17705      +66     
+ Misses       1816     1769      -47

Flag	Coverage Δ
cpp	`90.91% <68.62%> (+0.25%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
common	`85.81% <ø> (-0.18%)`	⬇️
datacell	`92.09% <ø> (+0.40%)`	⬆️
index	`90.58% <ø> (+0.66%)`	⬆️
simd	`100.00% <ø> (ø)`

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d091c9e...a533980. Read the comment docs.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

LHT129

LHTM

…1680) * Initial plan * Fix memory leak in FlattenDataCell query and ComputePairVectors on exception Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com> * fix(datacell): fix memory leak in FlattenDataCell query on exception Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com> --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com>

Initial plan

51b7d3d

Copilot AI assigned Copilot and wxyucs Mar 12, 2026

pull-request-size bot added the size/XS label Mar 12, 2026

Copilot started work on behalf of wxyucs March 12, 2026 02:33 View session

wxyucs added kind/bug Something isn't working version/1.0 needs-cherry-pick-release-0.18 labels Mar 12, 2026

Fix memory leak in FlattenDataCell query and ComputePairVectors on ex…

81978fd

…ception Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com>

pull-request-size bot added size/L and removed size/XS labels Mar 12, 2026

Copilot AI changed the title ~~[WIP] [HGRAPH-101] Fix HGraph memory leak~~ Fix memory leak in FlattenDataCell when exceptions occur during query Mar 12, 2026

Copilot finished work on behalf of wxyucs March 12, 2026 03:05

wxyucs marked this pull request as ready for review March 12, 2026 07:44

wxyucs requested a review from jiaweizone as a code owner March 12, 2026 07:44

Copilot AI review requested due to automatic review settings March 12, 2026 07:44

wxyucs requested review from LHT129, ShawnShawnYou, inabao and wxyucs as code owners March 12, 2026 07:44

Copilot started reviewing on behalf of wxyucs March 12, 2026 07:44 View session

Copilot started work on behalf of wxyucs March 12, 2026 07:45 View session

Copilot AI reviewed Mar 12, 2026

View reviewed changes

fix(datacell): fix memory leak in FlattenDataCell query on exception

a533980

Co-authored-by: wxyucs <12595343+wxyucs@users.noreply.github.com>

Copilot AI changed the title ~~Fix memory leak in FlattenDataCell when exceptions occur during query~~ fix(datacell): fix memory leak in FlattenDataCell query on exception Mar 12, 2026

Copilot finished work on behalf of wxyucs March 12, 2026 07:51

inabao approved these changes Mar 12, 2026

View reviewed changes

LHT129 approved these changes Mar 13, 2026

View reviewed changes

wxyucs merged commit bf48a2a into main Mar 13, 2026
23 checks passed

wxyucs deleted the copilot/fix-hgraph-memory-leak branch March 13, 2026 03:41

wxyucs added the cherry-pick-0.18-done label Mar 13, 2026

inabao added the needs-cherry-pick-release-0.16 label Mar 16, 2026

wxyucs added the cherry-pick-0.16-done label Mar 16, 2026

wxyucs added needs-cherry-pick-release-0.15 cherry-pick-0.15-done labels Mar 30, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(datacell): fix memory leak in FlattenDataCell query on exception#1680

fix(datacell): fix memory leak in FlattenDataCell query on exception#1680
wxyucs merged 3 commits intomainfrom
copilot/fix-hgraph-memory-leak

Copilot AI commented Mar 12, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 12, 2026

Uh oh!

inabao left a comment

Uh oh!

codecov bot commented Mar 12, 2026 •

edited

Loading

Uh oh!

LHT129 left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

Copilot AI commented Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Comments on the Issue (you are @copilot in this section)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Mar 12, 2026

Choose a reason for hiding this comment

Uh oh!

inabao left a comment

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Mar 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

LHT129 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Copilot AI commented Mar 12, 2026 •

edited

Loading

codecov bot commented Mar 12, 2026 •

edited

Loading