ci(mutation): Stryker.NET 5 with the MTP runner, shards sized by mutant count, zero-kill and missing-shard guards by dlrivada · Pull Request #1713 · dlrivada/Encina

dlrivada · 2026-10-03T14:41:40Z

Part of #1441 (phase 1), #1440 and #1682. The closing keywords, and the knowledge records for #1440 and #1682, are added in phase 2, once the CI validation run proves that mutants are killed and that a missing shard is reported.

Summary

Under xUnit v3 the VsTest runner of Stryker.NET 4.x killed 0 mutants (#1440), per-folder test-case-filter was the only thing keeping shards inside their timeouts, and shard 14 could lose its runner and upload nothing while the run stayed green (#1682). The Mutation Tests workflow now runs Stryker.NET 5.0.0 with the Microsoft Testing Platform (MTP) runner and fails loudly when a shard proves nothing.

Stryker 5.0.0 + MTP (.config/dotnet-tools.json, .github/stryker-config.json): test-runner: mtp, coverage-analysis: off, project mode (project: Encina.csproj) from tests/Encina.UnitTests. The solution, test-projects and test-case-filter keys are gone (MTP ignores the filter, stryker-net#3757). Absorbs Dependabot PR deps: Bump dotnet-stryker from 4.16.0 to 5.0.0 #1527.
run-stryker.cs: runs Stryker from tests/Encina.UnitTests; --log-to-file is off by default (under MTP it writes about 40 MB per test run per server, gigabytes for a 150-mutant shard; still available with -- --log-to-file); -c/--configuration is forwarded only when given (it was parsed and never passed before).
Shards by mutant count (mutation-tests.yml): a SHARDS array of ;-separated mutate globs (with ! exclusions) replaces FOLDERS/FILTERS. 20 shards of about 150 tested mutants or fewer, from run 36959365309 (Modules/Isolation from run 36304862790). A large folder is split by naming its biggest files; its last shard takes the folder glob minus those files, so a new file always lands in a shard. Modules.Isolation is four shards; the two permission-script generators (201 and 185 mutants) get one shard each.
Provisional TIMEOUTS from the [INFRA] Mutation Tests: 8 of 17 shards hit the 60-minute job limit, so the weekly run never completes or publishes #1395 formula (20 min setup + 35.6 s per mutant, measured locally): 81 to 209 minutes. Phase 2 re-sizes them from the validation run.
Build: shard jobs restore and build only Encina.UnitTests; test-baseline runs only Encina.UnitTests.
Guards: a matrix shard that tested mutants and killed none fails, and the aggregate refuses to publish ([INFRA] Mutation Tests: the Stryker VsTest runner kills 0 mutants under xUnit v3, so every mutation score is 0 % by construction #1440); a missing shard report raises a ::warning, a "Missing shard reports" summary block and a row in the table ([INFRA] Mutation shard 14 (Modules.Isolation) loses its hosted runner and uploads nothing, and the run stays green #1682); report and summary uploads run with if: always(); memory and disk go to runner-resources.log every minute and every fifth reading to the step log ([INFRA] Mutation shard 14 (Modules.Isolation) loses its hosted runner and uploads nothing, and the run stays green #1682). The fix(ci): mutation aggregate keeps every shard's mutants and fails on overlap or a count mismatch (#1681) #1684 overlap and count guards are kept. Permissions are declared per job (AGENTS.md §10); the workflow-level block is gone.
Docs: docs/testing/mutation-measurement-methodology.md, docs/en/guides/MUTATION_TESTING.md; knowledge record docs/knowledge/issues/1441.md (outcome partial). The AGENTS.md §9 mutation rule now describes SHARDS/TIMEOUTS under the MTP runner. Stale pages outside this diff: [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711.

Verification

Local smoke run (one file, SequentialDispatchStrategy.cs): 22,069 tests found, 4 mutants tested, 4 killed, score 100 %; 4 mutants in 92 s at concurrency 2.
Exclusion-glob probe: **/Dispatchers/Strategies/*.cs plus ! entries selected only the remaining file's 4 mutants (the remainder-shard design works).
actionlint clean on mutation-tests.yml; run-stryker.cs builds with no warnings.
Self-review (adversarial-reviewer): simulated the 20 shards against every file in src/Encina: no gap or overlap against the old 17 folders; counts and timeouts match. Four minors fixed (step-log readings, zero-kill guard only in matrix mode, configuration message, a docs claim).
Pending (phase 2, before merge): calibration run 37130291929 (custom_scope=**/Dispatchers/Strategies/*.cs) measures the CI time per mutant (local figure is from 32 cores; runners have 4 vCPU). Then the full 20-shard validation run, re-sized TIMEOUTS (max(60, ceil(minutes × 1.5))), and the record and docs updated with the observed figures.

Cross-cutting checklist (ADR-018)

Not applicable: CI and test tooling only; no entity, store, behavior, service or external integration in src/.

…de from Encina.UnitTests (#1441) Bump dotnet-stryker to 5.0.0, set test-runner mtp and project Encina.csproj in stryker-config.json (no solution, test-projects or test-case-filter keys), run Stryker from tests/Encina.UnitTests, and stop passing --log-to-file by default because the MTP runner's trace JSON-RPC logs grow by about 40 MB per test run.

…d missing and zero-kill shards (#1441, #1440, #1682) Replace FOLDERS/FILTERS with SHARDS of ';'-separated mutate globs (20 shards of at most about 150 tested mutants, large folders split by file with an exclusion-based remainder shard), set provisional TIMEOUTS from the #1395 formula, drop the inert test-case-filter patching, build only Encina.UnitTests, log runner memory and disk, upload reports even when a step fails, fail a shard and the aggregate when a report tested mutants but killed none, and warn in the summary when a shard report is missing.

…-count shards (#1441, #1440, #1682)

…hards, guards)

…l only in matrix mode, forward --configuration only when given (#1441)

…er-resources step log and optional --configuration (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

coderabbitai · 2026-10-03T14:41:48Z

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

Autopilot · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

…unner (#1441)

…one-element loop (#1441)

dlrivada · 2026-10-03T14:52:39Z

Review of PR #1713 (phase 1): merge after fixes and the validation run

Multi-dimension review with adversarial verification: each finding was checked by two independent skeptics. CONFIRMED = both found it real; PLAUSIBLE = one did. Refuted findings are listed at the end.

Confirmed (5)

nit [workflow] .github/workflows/mutation-tests.yml:236: Single-element for loop left in the diagnostic baseline. The baseline step keeps a bash for proj in \ tests/Encina.UnitTests/Encina.UnitTests.csproj; do loop around one project, with name=$(basename ...). It works, but it is leftover structure from the seven-project version. Fix: Replace the loop with a direct dotnet test call.
minor [shards] AGENTS.md:129: AGENTS.md section 9 still describes the removed FOLDERS/FILTERS arrays and the test-case-filter in stryker-config.json. AGENTS.md line 129 tells contributors to update the parallel FILTERS array when adding or renaming a folder, and says the test filter goes in stryker-config.json (test-case-filter). This PR deletes the FILTERS array, FOLDERS and the test-case-filter key (diff of mutation-tests.yml and .github/stryker-config.json), and mutation-measurement-methodology.md now says they are gone. AGENTS.md is not in the diff, so the binding rule now points at things that do not exist. The edit may belong to the orchestrator, since AGENTS.md is a shared hot spot. Fix: Replace the rule with: a new rotation folder is added as a SHARDS entry with a paired TIMEOUTS value, sized by tested-mutant count (target about 150 or fewer); there is no test filter under the MTP runner.
major [docs] docs/knowledge/issues/1441.md:1: PR says Fixes #1440 and Fixes #1682 but adds no knowledge records for them (pr-cycle REQ-031). .claude/skills/pr-cycle/SKILL.md:19 says a PR that closes an issue (Fixes #n) adds that issue's record docs/knowledge/issues/<n>.md in the same PR. The diff adds only 1441.md, and [INFRA] Mutation Tests: move Stryker to 5.0.0 with the MTP runner, coverage off, in project mode, and redesign the shards #1441 is a 'Part of' issue. No 1440.md or 1682.md exists. The two issues this PR does close have no record. Fix: Add records for [INFRA] Mutation Tests: the Stryker VsTest runner kills 0 mutants under xUnit v3, so every mutation score is 0 % by construction #1440 and [INFRA] Mutation shard 14 (Modules.Isolation) loses its hosted runner and uploads nothing, and the run stays green #1682 (valid, closed, with audit) in this PR, or in phase 2 together with the validation-run evidence. Alternatively drop the Fixes keywords until the validation run proves them, then add the records.
minor [docs] docs/testing/mutation-measurement-methodology.md:130: Guards section says the merge 'fails only when no shard uploaded a report', which contradicts the guards listed next to it. The 'Missing reports are named' bullet ends with 'The merge fails only when no shard uploaded a report.' The same page, and the workflow, say the aggregate also fails on a zero-kill report (mutation-tests.yml:596-600), on files with mutants in two reports (614-618) and on a merged-count mismatch (630-633). The sentence is true only for the missing-report case. Fix: Reword: 'A missing report alone does not fail the merge; it fails with "No shard reports found" only when no shard uploaded one.'
minor [docs] docs/engineering/REPLICATION-GUIDE.md:84: [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711 does not list every stale statement: REPLICATION-GUIDE.md:84 (test-case-filter in the config description) is missing. [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711 lists REPLICATION-GUIDE.md:115 only. Line 84 also describes .github/stryker-config.json as holding test-case-filter and a '17-shard ma[trix]', both now false. Within Stryker-xUnit-v3.md, [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711 names lines 15, 18, 28, 37, 55, but lines 31, 33, 38, 43, 63 and 72 also carry stale or version-pinned statements; the issue does cover the file as a whole in its Proposed Fix, so the fix is only a line-list gap. AGENTS.md:129 (the binding §9 rule telling contributors to maintain FILTERS and test-case-filter) stays wrong from merge until the orchestrator edits it. [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711 does record it, but its fix is gated 'after the validation run'. Fix: Add REPLICATION-GUIDE.md:84 to [DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711. Consider fixing AGENTS.md:129 in the same merge as this PR rather than after the validation run, since the workflow change makes the rule actively misleading.

Plausible (3)

minor [workflow] .github/workflows/mutation-tests.yml:22: Workflow-level permissions block (pre-existing on origin/main) contradicts AGENTS.md section 10. A top-level permissions: contents: read exists. It is not new: git show origin/main:.github/workflows/mutation-tests.yml has the same block. AGENTS.md section 10 says permissions are declared per job, never at workflow level. The PR rewrites most of this file and leaves the block in place. No job here needs more than contents: read, so the move is mechanical. Fix: Move permissions: contents: read into each of the select-matrix, test-baseline, run-mutation-tests and aggregate jobs, or record it as a known deviation.
nit [stryker] .github/scripts/run-stryker.cs:57: Pass-through arguments with relative paths now resolve against tests/Encina.UnitTests, undocumented. Stryker is now started with working directory tests/Encina.UnitTests (run-stryker.cs:57-59), but the script still resets its own cwd to the repo root. A relative path in a pass-through argument (for example --output, --reporter file paths, --baseline paths) resolves against the test project directory, not the repo root. --mutate globs are unaffected (relative to the mutated project) and the script passes absolute paths for --config-file and --output. The docs say to run the command from the repository root, which could suggest repo-root-relative paths. No current call site is affected. Fix: Add one sentence to MUTATION_TESTING.md and to the script comment saying that extra path arguments are relative to tests/Encina.UnitTests, or tell users to pass absolute paths.
minor [docs] docs/testing/mutation-measurement-methodology.md:100: Timing and kill figures from a local smoke run are hand-typed and unbacked; the PR body and docs disagree on its duration. The methodology page (Smoke check section) and MUTATION_TESTING.md state 'tested 4 mutants, killed all 4, in 8 min 41 s (initial test run about 4 min 40 s)'. The PR body says '4 mutants in 92 s at concurrency 2'. 8:41 minus 4:40 leaves about 241 s for the mutant phase, not 92 s. The figures are not reproducible from anything in the repo or CI (no artifact, no mutref marker), and 'killed 4 of 4' is a mutation result typed by hand (SPEC-001). The '22,069 tests' figure is also unbacked, and differs from the 21,091 of spike [SPIKE] Evaluate Stryker.NET 5.0.0 MTP runner with perTest coverage on one mutation shard #1087 without saying why. The page also states ubuntu-latest has '4 vCPU and 16 GB' with no source. Fix: Reconcile the two durations (state what 92 s measures) or drop it from the PR body. Label the figures as a dated local observation, with no score wording, or replace them with the calibration run's numbers once run 37130291929 completes. Cite a GitHub doc URL for the runner size.

Refuted (11)

[workflow] Provisional TIMEOUTS rest on a local figure that the repo's own data says is optimistic for CI (Class C)
[workflow] Resource monitor keeps the step's stdout open after the trap kills it
[workflow] Zero-kill guard is blind to a shard whose mutants are all RuntimeError, and is skipped in custom and diff mode
[shards] TIMEOUTS rest on a per-mutant figure that the spike itself marks as inflated by flaky kills, and on a 20-minute setup guess measured on a 32-core machine
[shards] The 20 shards still cover only 125 of the 285 mutable files of src/Encina; the PR text does not say so
[shards] Comment format for combined shards (5, 6, 11) is a sum expression, not a single count
[stryker] Dropping --log-to-file leaves CI with only info-level console output for failed shards
[docs] Knowledge record fails knowledge-records --check (11 errors); the required knowledge-records CI job is already red on this PR
[docs] Fixes #1440 and Fixes #1682 are not yet justified; the evidence is a pending run
[docs] Guide tells readers to 'run from the repository root' because the script runs Stryker from tests/Encina.UnitTests; the script already resolves the root itself
[docs] Timeouts section is provisional but the PR body cites 20 shards 'of about 150 tested mutants or fewer' while 3 shards exceed it

#1441)

…n merge (#1441)

…nitTests (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

Stryker 5.0.0 under-reports kills with the MTP runner above concurrency 1 (stryker-net#3832), and each slot keeps a multi-GB test host alive. The value lives in the config because run-stryker.cs reads -c as --configuration.

…tant merge for span-split files (#1441) The Run mutation tests step caps the managed heap at 4 GiB (DOTNET_GCHeapHardLimit), raises its own OOM score and runs Stryker under nice so a memory or CPU squeeze takes Stryker or a test host instead of the runner agent, and tees Stryker's console to stryker-console.log with pipefail. A monitor publishes runner readings every minute (every PUBLISH_EVERY minutes on the GitHub API, sized to the shard count) to a 'mutation telemetry (shard N)' check run that survives a lost runner; a final step and the aggregate job complete it. SHARDS can split one file with Stryker's character-offset span suffix; Build matrix rejects a span glob without its twin, and the aggregate merges reports mutant by mutant, drops empty and 'Removed by mutate filter' placeholders, and fails when two reports hold the same mutant. SHARDS and TIMEOUTS stay marked pending the phase-2 measurement.

… splits and per-mutant merge of the mutation workflow (#1441)

…, telemetry, span splits) (#1441)

…nd slower telemetry updates (#1441) A shard whose Stryker console reports an OutOfMemoryException fails and does not upload its report, since a test failing on the heap cap may count as a kill; the visible-failure effect of the cap is marked as a hypothesis for the calibration runs. The check-run token no longer reaches Stryker or the test hosts, every gh call times out, check-run updates stay near 360 per hour for the whole matrix, a failed oom_score_adj write only warns, and the readings also count processes naming Encina.UnitTests.

…ration; document the OOM guard, publish rate, token handling and process counts (#1441)

…fter review (#1441)

…lemetry rate, Debug build) (#1441)

… for custom scopes, restore the weekly schedule (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

…erver (#1441 phase 2f) Phase 2e measured that glibc's dynamic mmap threshold kept the native memory freed after each run of the ABAC EEL tests, so the reused test server grew ~2.5 GB per run. Fixed 128 KiB thresholds cut the first-run peak from 7.6 GB to 4.9 GB and the growth to ~230 MB per run with no slowdown.

…during Stryker runs (#1441 phase 2f) Adds an MTP TestingPlatformBuilderHook to Encina.UnitTests that registers an ITestSessionLifetimeHandler only when STRYKER_MUTANT_FILE is set. At the start of a test session after the first one in the process, it kills the test server when its private memory is above ENCINA_MTP_RECYCLE_MB (default 6144 MB) and writes one [encina-mtp-recycle] line to stderr. Stryker 5.0.0 treats the lost connection as a crashed host, discards the server and reruns the same mutant on a fresh one (RunAssemblyTestsInternalAsync, two attempts), so the verdict comes from the second attempt. The first session of a process never recycles, so the retry cannot fail because of the hook.

#1441 phase 2f) Review fixes for the MTP test server recycler: - Stryker 5.0.0 discards the test server's stderr unless --log-to-file is set, so the recycle line is also appended to the file named by ENCINA_MTP_RECYCLE_LOG when that variable is set. - ENCINA_MTP_RECYCLE_MB values too large to express in bytes are capped instead of overflowing into a negative threshold that recycles every session. - The builder hook moves to its own file and takes the environment reader as a parameter, so tests prove it registers nothing without STRYKER_MUTANT_FILE.

…ecycler (#1441)

…cle count in the shard summary (#1441 phase 2f)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

dlrivada · 2026-10-05T20:40:08Z

#1441 phase 2f, ready for verification. Branch head 28dbfad8.

Changes:

MALLOC_MMAP_THRESHOLD_ and MALLOC_TRIM_THRESHOLD_ are fixed at 128 KiB.
New test-server recycler, StrykerServerRecycler:
- it is active only when STRYKER_MUTANT_FILE is set;
- at the start of a mutant's session it kills its own process if private memory is above ENCINA_MTP_RECYCLE_MB (default 6144). It never does this on a process's first session;
- Stryker 5.0.0 then retries the mutant on a fresh server (MicrosoftTestingPlatformRunner.cs#L1184-L1244);
- 27 tests cover it.
Every recycle is logged to ENCINA_MTP_RECYCLE_LOG, uploaded with the stryker-logs artifact, and counted in the shard summary.

Local Windows runs with and without a recycle gave identical verdicts.

Verification run 37371127806 uses custom_scope **/Dispatchers/Strategies/*.cs, the same scope as the phase 2c/2d OOM runs. It is accepted when all of these hold:

no RuntimeError mutants;
every "failed on attempt 1/2" line is followed by a verdict;
the cgroup peak stays under 12 GiB;
the kill set matches a run without recycles.

Related: #1858 / PR #1860 (EEL tests share a static compiler) and #1859 (collectible ALC in src).

codecov · 2026-10-05T21:34:46Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

…p memory.events (#1441 phase 2f)

… id (#1441)

…cle threshold and cg_ev_* telemetry (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

…ance timeout margin (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

dlrivada · 2026-10-06T13:38:06Z

Follow-up for the rewritten docs/testing/mutation-measurement-methodology.md: #1900 adds the job-graph diagram once this PR merges (the page still has no Mermaid block).

…arallel strategies directly (#1441)

Copilot

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

dlrivada added 6 commits October 3, 2026 16:16

docs(testing): document Stryker 5 MTP runner, project mode and mutant…

7518930

…-count shards (#1441, #1440, #1682)

docs(knowledge): record for #1441 phase 1 (MTP runner, mutant-count s…

e662762

…hards, guards)

ci(mutation): stream runner resources to the step log, guard zero-kil…

743464f

…l only in matrix mode, forward --configuration only when given (#1441)

docs(testing): mutation docs follow matrix-only zero-kill guard, runn…

598d9b2

…er-resources step log and optional --configuration (#1441)

Copilot AI balanced review requested due to automatic review settings October 3, 2026 14:41

Copilot AI reviewed Oct 3, 2026

docs(knowledge): #1441 record follows the knowledge-record schema

6faa05f

Copilot AI balanced review requested due to automatic review settings October 3, 2026 14:46

Copilot AI reviewed Oct 3, 2026

dlrivada added 2 commits October 3, 2026 16:52

docs(agents): mutation rule describes SHARDS/TIMEOUTS under the MTP r…

5d711c9

…unner (#1441)

ci(mutation): run the diagnostic baseline test directly instead of a …

5c63ae3

…one-element loop (#1441)

dlrivada mentioned this pull request Oct 3, 2026

[DEBT] Mutation docs and rules outside the methodology page still describe the VsTest runner, per-folder test filters and the 17-shard matrix #1711

Open

3 of 14 tasks

dlrivada added 3 commits October 3, 2026 16:53

ci(mutation): declare permissions per job instead of at workflow level (

be6620f

#1441)

docs(testing): a missing shard report alone does not fail the mutatio…

a27d9e6

…n merge (#1441)

docs(testing): extra Stryker arguments resolve against tests/Encina.U…

6e3bf5c

…nitTests (#1441)

Copilot AI balanced review requested due to automatic review settings October 3, 2026 14:53

Copilot AI reviewed Oct 3, 2026

dlrivada added 8 commits October 3, 2026 21:21

ci(mutation): run Stryker at concurrency 1 (#1441)

1467934

Stryker 5.0.0 under-reports kills with the MTP runner above concurrency 1 (stryker-net#3832), and each slot keeps a multi-GB test host alive. The value lives in the config because run-stryker.cs reads -c as --configuration.

docs(testing): describe concurrency 1, GC cap, runner telemetry, span…

4837e3d

… splits and per-mutant merge of the mutation workflow (#1441)

docs(knowledge): #1441 record covers phase 2 (concurrency 1, heap cap…

2ec662b

…, telemetry, span splits) (#1441)

docs(testing): mutation runner protection is a hypothesis until calib…

cd5f498

…ration; document the OOM guard, publish rate, token handling and process counts (#1441)

docs(testing): qualify the report-upload guard and the update count a…

629f23c

…fter review (#1441)

docs(knowledge): #1441 record follows the review fixes (OOM guard, te…

e260d33

…lemetry rate, Debug build) (#1441)

Copilot AI balanced review requested due to automatic review settings October 3, 2026 19:51

dlrivada added 3 commits October 5, 2026 12:42

Merge remote-tracking branch 'origin/main' into ci/stryker5-mtp-1441

f109429

ci(mutation): force workstation GC for the MTP test server, debug log…

727f669

… for custom scopes, restore the weekly schedule (#1441)

docs(knowledge): #1441 phase 2c diagnosis stated as unproven

ce9660d

Copilot AI balanced review requested due to automatic review settings October 5, 2026 10:50

Copilot AI reviewed Oct 5, 2026

docs(mutation): #1441 phase 2e native memory findings

76fb24d

dlrivada mentioned this pull request Oct 5, 2026

[DEBT] EEL unit tests build a fresh EELCompiler per test, so every run re-emits Roslyn script assemblies the process never unloads (MTP test-server growth under Stryker) #1858

Closed

4 of 14 tasks

dlrivada added 5 commits October 5, 2026 21:17

docs(mutation): record the phase 2f plan status and the test server r…

72711e3

…ecycler (#1441)

ci(mutation): recycle threshold and log for the MTP test server, recy…

28dbfad

…cle count in the shard summary (#1441 phase 2f)

Copilot AI balanced review requested due to automatic review settings October 5, 2026 20:39

Copilot AI reviewed Oct 5, 2026

dlrivada added 3 commits October 6, 2026 12:18

ci(mutation): recycle the MTP test server at 4096 MB and sample cgrou…

3cbefa8

…p memory.events (#1441 phase 2f)

ci(mutation): word the 4096 MB rationale as a hypothesis with the run…

0db4001

… id (#1441)

docs(mutation): record the phase 2f CI verification, the 4096 MB recy…

ecce598

…cle threshold and cg_ev_* telemetry (#1441)

Copilot AI balanced review requested due to automatic review settings October 6, 2026 11:07

Copilot AI reviewed Oct 6, 2026

ci(mutation): recycle at 4608 MB and give custom scopes a runner-vari…

32d3650

…ance timeout margin (#1441)

Copilot AI balanced review requested due to automatic review settings October 6, 2026 13:17

Copilot AI reviewed Oct 6, 2026

dlrivada mentioned this pull request Oct 6, 2026

[DEBT] The mutation methodology page explains the pipeline in prose only: add the job-graph diagram after #1713 merges #1900

Open

3 of 14 tasks

test(dispatch): kill the conditional and initializer mutants of the p…

2a56f12

…arallel strategies directly (#1441)

Copilot AI balanced review requested due to automatic review settings October 6, 2026 15:46

Copilot AI reviewed Oct 6, 2026

dlrivada mentioned this pull request Oct 6, 2026

[DEBT] Method docs drift: ai-task-routing subagent table out of sync with .claude/agents, stale project count in HOW-ENCINA-IS-BUILT #1907

Open

3 of 14 tasks

Conversation

dlrivada commented Oct 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Verification

Cross-cutting checklist (ADR-018)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot commented Oct 3, 2026

Draft PR not reviewed

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

dlrivada commented Oct 3, 2026

Review of PR #1713 (phase 1): merge after fixes and the validation run

Confirmed (5)

Plausible (3)

Refuted (11)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

dlrivada commented Oct 5, 2026

Uh oh!

codecov Bot commented Oct 5, 2026

Codecov Report

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

dlrivada commented Oct 6, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

dlrivada commented Oct 3, 2026 •

edited

Loading