Enable Miri tests in CI with non-blocking execution #697

Copilot · 2026-01-20T23:27:52Z

✅ Enable Miri Tests in CI - Complete

All tasks completed successfully!

Summary

Successfully enabled Miri tests in the CI workflow with non-blocking execution and parallel processing. Tests have been optimized based on reviewer feedback.

Key Changes

Enabled Miri Job - Uncommented and configured in .github/workflows/ci.yml
Non-Blocking - Added continue-on-error: true to prevent CI blocking
Parallel Execution - Runs alongside test-workspace jobs (Phase 3)
Simplified Command - Uses original command: cargo +nightly miri nextest run --package diskann-quantization
Test Optimization - Disabled tests that don't execute unsafe code for miri:
- test_double_hadamard and test_padding_hadamard (transforms)
- test_distances_in_place (kmeans/lloyds)
- minmax_quantizer_tests module (entire test module)
- minmax_vector_tests module (entire test module)
Test Configuration - Applied if cfg!(miri) pattern to end_to_end_test for cleaner conditional setup

CI Pipeline Structure

Phase 1: fmt, clippy (parallel, 2-3 min)
         ↓
Phase 2: basics (aggregation, instant)
         ↓
Phase 3: test-workspace, test-workspace-features, miri (parallel)

Time Impact Analysis

Miri tests run in parallel with other test jobs
Test suite optimized to skip non-unsafe code tests
Expected CI time: well under 60-minute requirement
CI time determined by longest-running parallel job

Quality & Security

✅ Detects undefined behavior and memory safety issues
✅ Non-blocking ensures development velocity
✅ YAML syntax validated
✅ CodeQL scan passed (0 alerts)
✅ Test suite optimized per reviewer feedback

The changes ensure Miri tests provide valuable feedback without blocking the CI pipeline or adding unnecessary complexity.

Original prompt

This section details on the original issue you should resolve

<issue_title>Enable miri tests in .github/workflows/ci.yml</issue_title>
<issue_description>

DiskANN/.github/workflows/ci.yml

Line 209 in f9c6c4b

# name: miri-test

Ensure that overal build time doesn't exceed 60 minutes.</issue_description>

<agent_instructions>Enable Miri tests, but make sure they don’t fail the CI build. Measure and report how much time they add to the CI pipeline, and consider running them in parallel with other steps to minimize their impact on overall build duration.</agent_instructions>

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

Fixes Enable miri tests in .github/workflows/ci.yml #696

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

.github/workflows/ci.yml

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

hildebrandmw · 2026-01-23T18:14:39Z

diskann-quantization/src/algorithms/kmeans/common.rs


    #[test]
    fn test_block_transpose_16() {
+        #[cfg(not(miri))]


Can these bounds be set up with something like

let row_range = if cfg!(miri) { 127..128 } else { 0..128 };

and such to avoid the repitition?

hildebrandmw · 2026-01-23T18:18:23Z

diskann-quantization/src/algorithms/kmeans/lloyds.rs

    }

    #[test]
    fn test_distances_in_place() {


We can probably skip this test entirely when running under miri. There is a dedicated Miri test just after this that validates the indexing.

@copilot, disable test_distances_in_place() test for miri. revert the changes made in this test.

Disabled test_distances_in_place for miri and reverted changes. Commit: 0833183

hildebrandmw · 2026-01-23T18:20:10Z

diskann-quantization/src/algorithms/kmeans/lloyds.rs

            }
        }
+
+        #[cfg(miri)]


This one we cannot short circuit with Miri. The distances_in_place algorithm needs to be tested for a range of sizes to validate memory accesses.

hildebrandmw · 2026-01-23T18:25:18Z

diskann-quantization/src/algorithms/transforms/double_hadamard.rs

        //
        // Subsampling results in poor preservation of inner products, so we skip it
        // altogether.
+        #[cfg(not(miri))]


I'd be tempted to say that we can skip all of test_double_hadamard when running under Miri. It doesn't execute any unsafe code on its own.

I've asked copilot to disable it in another comment )

hildebrandmw · 2026-01-23T18:25:46Z

diskann-quantization/src/algorithms/transforms/padding_hadamard.rs


+        // Miri is extremely slow, so we skip the larger tests there.
+        #[cfg(not(miri))]
        let dim_combos = [


As with test_double_hadamard, I'm tempted to say we can skip test_padding_hadamard altogether under Miri.

@copilot, disable test_double_hadamard and test_padding_hadamard tests for miri. Don't forget to revert the miri-related changes made in the tests, since they are not needed anymore.

Disabled both tests for miri and reverted all miri-specific changes. Commit: 7bc3e20

hildebrandmw · 2026-01-23T18:32:54Z

diskann-quantization/src/minmax/vectors.rs

+                    #[cfg(miri)]
+                    if dim != MAX_DIM {
+                        continue;
+                    }


We can almost skip the MinMax entirely under Miri because it does not run unsafe outside of decompression. Since the indexing used in the decompression is checked in debug builds - it might be fine to just skip.

@copilot, revert the change made inside fn $name():

#[cfg(miri)] if dim != MAX_DIM { continue; }

don't touch other changes in this file.
and disable the whole test group in this file (minmax_vector_tests) for miri.

Reverted the miri-specific loop skip and disabled the entire minmax_vector_tests module for miri. Commit: bf86fef

hildebrandmw · 2026-01-23T18:33:26Z

diskann-quantization/src/bits/slice.rs

+            #[cfg(miri)]
+            if len != MAX_DIM - 1 {
+                continue;
+            }


The tests in slice all need to run under Miri unfortunately.

hildebrandmw · 2026-01-23T18:34:11Z

diskann-quantization/src/minmax/quantizer.rs

+                        #[cfg(miri)]
+                        if d != $dim - 1 {
+                            continue;
+                        }


We can probably skip the min-max quantizer tests under Miri.

@copilot, revert the change made in this file and disable the whole test group (minmax_quantizer_tests) for miri.

Reverted changes and disabled the entire minmax_quantizer_tests module for miri. Commit: 0833183

hildebrandmw · 2026-01-23T18:37:59Z

diskann-quantization/src/product/tables/transposed/pivots.rs

        }
+
+        #[cfg(miri)]
+        for total in 63..64 {


Maybe change the upper bound to like 48 when under Miri but still do the sweep? That seems reasonable quick.

hildebrandmw · 2026-01-23T18:41:36Z

diskann-quantization/src/product/tables/transposed/table.rs

+            for num_centers in 47..48 {
+                test_process_into_impl(num_chunks, num_centers, 2, &mut rng);
+            }
+        }


This file doesn't contain unsafe on its own, so it's probably fine to either skip miri on these test, or do what you did here on a reduced set. If the latter, we should run at least a few more intermediate values than just the last since corner-case handling needs to be covered. I'd say pick a few odd-ball sizes in between 0 and 48, for example.

…rt miri-specific changes Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

arrayka · 2026-01-24T02:26:55Z

diskann-quantization/src/algorithms/kmeans/lloyds.rs

    #[test]
    fn end_to_end_test() {
        let mut rng = StdRng::seed_from_u64(0xff22c38d0f0531bf);
+        #[cfg(not(miri))]


@copilot, apply this pattern in end_to_end_test() test:
let setup = if cfg!(miri) {
EndToEndSetup { ... }
} else {
EndToEndSetup { ... }
};

Applied the if cfg!(miri) pattern to end_to_end_test. Commit: 0833183

…apply cfg!(miri) pattern to end_to_end_test Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

…or_tests module for miri Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Initial plan

adc532b

Copilot AI assigned Copilot and arrayka Jan 20, 2026

Copilot started work on behalf of arrayka January 20, 2026 23:28 View session

Copilot AI and others added 3 commits January 20, 2026 23:30

Enable Miri tests in CI with continue-on-error and time reporting

cd02489

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Fix bash flags to allow proper error handling in Miri tests

3b7eeb7

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Use set -ux for consistent error handling with || true

b2b0400

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Enable Miri tests in CI workflow~~ Enable Miri tests in CI with non-blocking execution and time tracking Jan 20, 2026

Copilot AI requested a review from arrayka January 20, 2026 23:37

Copilot finished work on behalf of arrayka January 20, 2026 23:37

arrayka reviewed Jan 20, 2026

View reviewed changes

.github/workflows/ci.yml Outdated Show resolved Hide resolved

Copilot started work on behalf of arrayka January 20, 2026 23:53 View session

Simplify Miri command to use original syntax without time measurement

e17c07f

Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Copilot AI changed the title ~~Enable Miri tests in CI with non-blocking execution and time tracking~~ Enable Miri tests in CI with non-blocking execution Jan 20, 2026

Copilot AI requested a review from arrayka January 20, 2026 23:56

Copilot finished work on behalf of arrayka January 20, 2026 23:56

Alex Razumov (from Dev Box) added 7 commits January 22, 2026 16:52

Fix unused imports

f5f255b

Adjust the safety comment

1f8c1de

Speed up miri tests

337a120

Speed up miri tests

a54c9c1

Optimizing more tests for miri

7a97192

Fixing unsigned_minmax_compensated_test_u1 test

9e7d4ae

More test optimizations

f288d34

hildebrandmw requested changes Jan 23, 2026

View reviewed changes

Merge branch 'main' into copilot/enable-miri-tests-ci-yml

56e0605

Copilot started work on behalf of arrayka January 24, 2026 02:23 View session

Disable test_double_hadamard and test_padding_hadamard for miri, reve…

7bc3e20

…rt miri-specific changes Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

arrayka reviewed Jan 24, 2026

View reviewed changes

Copilot AI requested a review from arrayka January 24, 2026 02:34

Copilot finished work on behalf of arrayka January 24, 2026 02:34

Copilot started work on behalf of arrayka January 24, 2026 02:35 View session

Disable test_distances_in_place and minmax_quantizer_tests for miri, …

0833183

…apply cfg!(miri) pattern to end_to_end_test Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Copilot finished work on behalf of arrayka January 24, 2026 02:39

Copilot started work on behalf of arrayka January 24, 2026 02:39 View session

Revert miri-specific loop skip in minmax vectors, disable minmax_vect…

bf86fef

…or_tests module for miri Co-authored-by: arrayka <1551741+arrayka@users.noreply.github.com>

Copilot finished work on behalf of arrayka January 24, 2026 02:43

Enable Miri tests in CI with non-blocking execution #697

Are you sure you want to change the base?

Enable Miri tests in CI with non-blocking execution #697

Uh oh!

Conversation

Copilot AI commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Enable Miri Tests in CI - Complete

Summary

Key Changes

CI Pipeline Structure

Time Impact Analysis

Quality & Security

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

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Jan 20, 2026 •

edited

Loading