test(rccl): add host microtests for net_ib_cast init, capability and GIN by aduforest · Pull Request #13840 · ROCm/rocm-systems

aduforest · 2026-10-20T19:17:09Z

JIRA ID: AICOMRCCL-2813

What

Host-only microtests for src/transport/net_ib_cast/{init,common,capability,reg,gdr,multiplane,gin,net_ib_cast_gid_inspect}.cc, in a new rccl-UnitTestsMicroNetIbCastInit with one test TU per unit.

How

The verbs, mlx5/ionic, socket and neighbouring net_ib_cast seams are shared fakes. The ones the other net_ib_cast PRs also use are byte-identical copies; this PR owns the scheduler, p2p_resiliency_recovery, gdr_peermem and gdr_probe fakes. Fake ibv contexts route post_send/poll_cq back to the test. xml.cc, utils.cc and net_telemetry.cc are the real ones. The GDR answers latched by std::call_once are tested in isolated processes.
transport_stubs.cc gains net.cc's NET_OPTIONAL_RECV_COMPLETION param, and os_fakes.cc gains ncclOsStrSep, the same hunk as test(rccl): add host microtests for src/bootstrap.cc #13601.
Pins 10 production bugs as KnownBug_ tests:
- AINIC fuses NICs by default, although the code means to require NCCL_IB_MERGE_NICS=1 (AICOMRCCL-3088).
- A GDR probe that could not run is latched as "no GDR" (AICOMRCCL-3089).
- A data-direct lookup error skips the cleanup (AICOMRCCL-3090).
- A failed device init turns the next one into a silent no-op (AICOMRCCL-3091).
- Deregistering unsorts the MR cache, so a cached range gets registered twice (AICOMRCCL-3092).
- A regMr failure on a later device keeps the earlier devices' MRs (AICOMRCCL-3093).
- GIN connect, context and registration failures leak comms and MRs (AICOMRCCL-3094).
- The RMA proxy's error report indexes devices by completion slot, and can crash (AICOMRCCL-3095).
- IPutSignal with data but no destination posts an empty WR instead of the signal (AICOMRCCL-3096).
- A plane-index failure frees the ibv device list twice (AICOMRCCL-3143).
Not pinned, as the outcome depends on uninitialized memory: IbCastMakeVDevice compares and logs PCI roots it never set when a path has none (AICOMRCCL-3097).

Coverage

Lines: init.cc 642/652, common.cc 185/185, common_cast.h 57/57, capability.cc 162/162, reg.cc 116/117, gdr.cc 73/73, multiplane.cc 186/189, gin.cc 491/491 (the GDAKI arm is compiled out, as in the shipping build), net_ib_cast_gid_inspect.cc 102/102. The misses are calloc failures, a capability probe that cannot fail, a telemetry slot that is never out of range, and the return after reg.cc's loop.

Mutation Testing

172 mutations on the hipified sources, all killed; 5 of them needed a sharper test first. 1 more against the review changes, also killed. 11 more for #12749's runtime uDMA count and its mlx5-only data-direct checks, all killed.

Test Plan

Standalone build: rccl-UnitTestsMicroNetIbCastInit 275/275 in about 0.4 s, and --gtest_shuffle --gtest_repeat=20 is clean. Every other host binary was rebuilt against the changed fakes and run once. DdaColl and Rocmwrap fail as they do on develop (the break #13819 fixes) and link none of the changed files.

Seams for scheduler.cc, p2p_resiliency_recovery.cc, gdr_peermem.cc and gdr_probe.cc, and transport_stubs gains net.cc's NET_OPTIONAL_RECV_COMPLETION param, for net_ib_cast microtest binaries that do not compile those TUs. Co-Authored-By: Claude <noreply@anthropic.com>

…seAinic Co-Authored-By: Claude <noreply@anthropic.com>

Byte-identical copies of the fakes the other net_ib_cast microtest PRs own (verbs, mlx5/ionic direct verbs, sockets, and the connect, p2p, p2p_resiliency and qp_sharing seams), and os_fakes' ncclOsStrSep as #13601 adds it. Co-Authored-By: Claude <noreply@anthropic.com>

rccl-UnitTestsMicroNetIbCastInit covers src/transport/net_ib_cast/{init,common, capability,reg,gdr,multiplane,gin,net_ib_cast_gid_inspect}.cc over fake verbs contexts, with the real xml.cc, utils.cc and net_telemetry.cc. Production bugs are pinned as KnownBug_ tests (AICOMRCCL-3088..3096). Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-20T19:17:14Z

@argus review

therock-pr-bot · 2026-10-20T19:17:59Z

✅ All Policy Checks Passed

Check	Status	Details
📝 PR Description	✅ Pass	—
⛔ Forbidden Files	✅ Pass	—
🧪 Unit Test	✅ Pass	—
🚫 Draft PR	🔜 To Be Enabled	—
🚩 Feature Flag	🔜 To Be Enabled	—
📊 Code Coverage	🔜 To Be Enabled	—

🎉 All policy checks passed!

📖 Need help? See the Policy FAQ for details on every check and how to fix failures.

🙋 Wish to Override Policy?

amd-argus

ARGUS

Findings

5 more finding(s), not posted inline

projects/rccl/test/host/net-ib-cast-multiplane-test.cc:191 (Low) Load_PipWithoutInterface_LeavesTheInterfaceEmpty checks only INFO log text, not the stored pip interface
projects/rccl/test/host/net-ib-cast-init-test.cc:222 (Low) GetRealPort test stores stack arrays in IbCastDevs[].pciPath; a failed ASSERT makes TearDown free() them and abort the binary
projects/rccl/test/host/net-ib-cast-gin-test.cc:1131 (Low) KnownBug death test expects KilledBySignal(SIGSEGV), which does not match under BUILD_ADDRESS_SANITIZER (ASan exits 1)
projects/rccl/test/host/fakes/net_ib_cast_qp_sharing_fakes.cc:15 (Low) Comment says CMakeLists.txt pins these RCCL_PARAM copies to qp_sharing.cc, but no rccl_assert_source_line covers them
projects/rccl/test/host/net-ib-cast-test-common.h:221 (Low) Control-statement bodies on their own line without braces (style guide requires braces); 6 sites

This PR changes 6,549 lines across 41 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

Pin the copied connect, qp_sharing, p2p, p2p_resiliency and socket fake defaults to production, check the stored PIP interface, let the slot-index crash pin accept ASan's exit, own the GetRealPort paths, and brace the single-statement bodies. Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-20T20:28:32Z

Addressed the five non-inline findings in 4d3155b:

multiplane: Load_PipWithoutInterface_LeavesTheInterfaceEmpty now also checks the stored PIP's interface is empty.
init: the GetRealPort test strdup()s its paths, so TearDown's free() is valid even after a failed ASSERT.
gin: the slot-index crash pin takes either SIGSEGV or ASan's non-zero exit.
qp_sharing fakes: test/host/CMakeLists.txt now pins the copied qp_sharing params (and the connect, p2p, p2p_resiliency and socket copies) to production with rccl_assert_source_line.
net-ib-cast-test-common.h and the tests: every single-statement body is braced.

aduforest · 2026-10-20T20:28:35Z

@argus re-review

…init-host-microtests # Conflicts: # projects/rccl/tools/scripts/test_runner/lib/test_executor.py

aduforest · 2026-10-20T20:53:32Z

@argus re-review

aduforest · 2026-10-20T23:54:34Z

@argus review

amd-argus

ARGUS

This PR changes 6,610 lines across 41 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-21T02:45:10Z

@argus review

amd-argus

ARGUS

This PR changes 6,610 lines across 41 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: Worth settling before approval: InitDevices_IbDisabled_FailsBeforeLoadingVerbs cannot fail as written, and the GIN error-path pins leak gin.cc allocations that LeakSanitizer reports in ASan builds.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

…_DISABLE test bite Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-21T05:12:54Z

Addressed both Medium findings in 5b15c49:

net-ib-cast-init-test.cc: InitDevices_IbDisabled_FailsBeforeLoadingVerbs now builds World() (with an HCA) before Init(); removing the IB_DISABLE check in init.cc now fails it.
net-ib-cast-gin-test.cc: the 8 tests that take a leaking gin.cc error path skip under ASan (kAsan, as in TopoScene.h). Checked with an ASan build: 264 pass, 8 skipped, LeakSanitizer clean (was 21 leaked allocations).

aduforest · 2026-10-21T05:12:56Z

@argus re-review

amd-argus

ARGUS

This PR changes 6,635 lines across 41 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: Worth settling before approval: InitDevices_NoVerbsLibrary_Fails passes with the verbs check removed, and the VDevice fixtures run strcmp on uninitialized PCI roots.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

…l PCI roots Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-21T06:06:48Z

Addressed both Medium findings in a7e2c2e:

InitDevices_NoVerbsLibrary_Fails builds World() first and asserts ionicSymbols.calls == 0; deleting the ibv check in init.cc now fails it (was a survivor).
The VDevice fixture's devices and the merged-device properties test get a shared pci0000:00 pciPath (no numa_node behind it), so the merge checks compare real roots. The NUMA/root-complex test sets its paths through the same helper, which frees the fixture's.

272/272 pass, --gtest_shuffle --gtest_repeat=20 clean, ASan build clean (264 pass, 8 skipped).

aduforest · 2026-10-21T06:06:50Z

@argus re-review

amd-argus

ARGUS

This PR changes 6,648 lines across 41 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

…comrccl-2813-net-ib-cast-init-host-microtests

…-1] read (pre-merge hardening) - ZeroedPtr allocates with the comm structs' alignas() alignment instead of calloc's 16 bytes, which cleared the UBSan misaligned-access reports in rccl-UnitTestsMicroNetIbCastInit. - The unknown-event case uses IBV_EVENT_DEVICE_SPEED_CHANGE, a real event the loop has no case for, instead of casting 999 into the enum. - MakeVDevice's negative-index case puts -1 second, since production reads IbCastDevs[devs[0]] before its range check. Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-22T05:56:18Z

@argus review

amd-argus

ARGUS

This PR changes 6,617 lines across 47 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

…comrccl-2813-net-ib-cast-init-host-microtests

Define ncclOsStrSep right after ncclOsGetPid, where #13840 adds the same fake, so merging the two branches keeps one definition instead of silently compiling two. Co-authored-by: Cursor <cursoragent@cursor.com>

…unt (pre-merge hardening) #12749 made init.cc call wrap_ionicdv_ctx_get_udma_count, so the Init binary no longer linked. ionicdvwrap_fakes gains hooks for it and for wrap_ionicdv_qp_get_udma_idx, defaulting to what the real wrappers return without libionic. Tests cover the normalized per-device count and the mlx5-only data-direct checks in GetProperties and DMA-BUF registration. Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-22T19:47:54Z

The remaining fake moves overlap Dhana's PRs, we agreed to merge them into one follow-up once both land

…comrccl-2813-net-ib-cast-init-host-microtests

Back to the fakes of 542196d: no socket_fakes, debug_fakes or ParamOverride copies, which conflicted with #13834, #13836 and #13840. Only the Won't Do rename of the AICOMRCCL-2966 pin stays. Co-Authored-By: Claude <noreply@anthropic.com>

Back to the content of 5209585 plus develop: the in-memory socket layer stays in bootstrap_micro_fakes.cc, which conflicted less with #13834, #13836 and #13840 than a socket_fakes copy. Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-22T20:07:49Z

@argus review

amd-argus

ARGUS

This PR changes 6,682 lines across 47 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: Worth settling before approval: InitDevices_TooManyPlanes_FailsAfterListing runs through a production double free of the device list (init.cc:591, then 740) without pinning it, and the telemetry and SendCommInit tests still pass with the code they check removed.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

Co-Authored-By: Claude <noreply@anthropic.com>

…comrccl-2813-net-ib-cast-init-host-microtests

aduforest · 2026-10-22T21:21:20Z

@argus re-review

amd-argus

ARGUS

This PR changes 6,721 lines across 47 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

Co-Authored-By: Claude <noreply@anthropic.com>

aduforest · 2026-10-22T21:48:56Z

@argus re-review

amd-argus

ARGUS

This PR changes 6,721 lines across 47 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

…comrccl-2813-net-ib-cast-init-host-microtests

…view Co-Authored-By: Claude <noreply@anthropic.com>

Host CI runs these binaries through run_host_tests.sh, so the json entries are not needed. Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus

ARGUS

This PR changes 6,725 lines across 47 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

aduforest · 2026-10-24T02:07:42Z

@argus review

amd-argus

ARGUS

This PR changes 6,708 lines across 46 files. A change this size is hard to review
well, for a person and for this bot alike: reviewers lose the thread, defects hide in the
volume, and the feedback arrives too late to be cheap to act on. Splitting it into a
stacked pull request
gives each piece a reviewable diff and keeps the history. This review covers the whole change
regardless.

Bottom line: LGTM.

_{Comment @argus review to re-run, or @argus re-review after pushing.}

github-actions · 2026-10-24T04:03:06Z

pr-rocm-systems-13840-oci-rccl-pr-smoke

Builds

Stack	Component	Version	Commit	Status
pr-13840_rccl-debug_rocm-10.1-ga	gcc-13.2	13.2.0	13.2.0	cached
pr-13840_rccl-debug_rocm-10.1-ga	rocm-10.1-ga	10.1.0	10.1.0	cached
pr-13840_rccl-debug_rocm-10.1-ga	openmpi-5.0.8	5.0.8	5.0.8	cached
pr-13840_rccl-debug_rocm-10.1-ga	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-debug_rocm-10.1-ga	rccl_gtest-pr-13840-debug	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.1-ga	gcc-13.2	13.2.0	13.2.0	cached
pr-13840_rccl-perf_rocm-10.1-ga	rocm-10.1-ga	10.1.0	10.1.0	cached
pr-13840_rccl-perf_rocm-10.1-ga	openmpi-5.0.8	5.0.8	5.0.8	cached
pr-13840_rccl-perf_rocm-10.1-ga	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.1-ga	rccl-tests-develop-perf	develop	`dc657f2`	cached
pr-13840_rccl-perf_rocm-10.2-latest	gcc-13.2	13.2.0	13.2.0	cached
pr-13840_rccl-perf_rocm-10.2-latest	rocm-10.2-latest	10.2.0rc0	10.2.0rc0	cached
pr-13840_rccl-perf_rocm-10.2-latest	openmpi-5.0.8	5.0.8	5.0.8	cached
pr-13840_rccl-perf_rocm-10.2-latest	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.2-latest	rccl-tests-develop-perf	develop	`dc657f2`	cached
rocm-10.1-ga_rccl-develop-perf	gcc-13.2	13.2.0	13.2.0	cached
rocm-10.1-ga_rccl-develop-perf	rocm-10.1-ga	10.1.0	10.1.0	cached
rocm-10.1-ga_rccl-develop-perf	openmpi-5.0.8	5.0.8	5.0.8	cached
rocm-10.1-ga_rccl-develop-perf	rccl-develop	develop	`dc657f2`	cached
rocm-10.1-ga_rccl-develop-perf	rccl-tests-develop-perf	develop	`dc657f2`	cached

Failures

No failing tests in this run.

15 tests skipped.

github-actions · 2026-10-24T04:03:07Z

pr-rocm-systems-13840-ruby64-rccl-pr-smoke

Builds

Stack	Component	Version	Commit	Status
pr-13840_rccl-debug_rocm-10.1-ga	gcc-11.5	11.5.0	11.5.0	cached
pr-13840_rccl-debug_rocm-10.1-ga	rocm-10.1-ga	10.1.0	10.1.0	cached
pr-13840_rccl-debug_rocm-10.1-ga	openmpi-4.0.2	4.0.2rc3	4.0.2rc3	cached
pr-13840_rccl-debug_rocm-10.1-ga	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-debug_rocm-10.1-ga	rccl_gtest-pr-13840-debug	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.1-ga	gcc-11.5	11.5.0	11.5.0	cached
pr-13840_rccl-perf_rocm-10.1-ga	rocm-10.1-ga	10.1.0	10.1.0	cached
pr-13840_rccl-perf_rocm-10.1-ga	openmpi-4.0.2	4.0.2rc3	4.0.2rc3	cached
pr-13840_rccl-perf_rocm-10.1-ga	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.1-ga	rcclt-develop-perf	develop	`dc657f2`	cached
pr-13840_rccl-perf_rocm-10.2-latest	gcc-11.5	11.5.0	11.5.0	cached
pr-13840_rccl-perf_rocm-10.2-latest	rocm-10.2-latest	10.2.0rc0	10.2.0rc0	cached
pr-13840_rccl-perf_rocm-10.2-latest	openmpi-4.0.2	4.0.2rc3	4.0.2rc3	cached
pr-13840_rccl-perf_rocm-10.2-latest	rccl-pr-13840	`62483ad`	`62483ad`	cached
pr-13840_rccl-perf_rocm-10.2-latest	rcclt-develop-perf	develop	`dc657f2`	cached
rocm-10.1-ga_rccl-develop-perf	gcc-11.5	11.5.0	11.5.0	cached
rocm-10.1-ga_rccl-develop-perf	rocm-10.1-ga	10.1.0	10.1.0	cached
rocm-10.1-ga_rccl-develop-perf	openmpi-4.0.2	4.0.2rc3	4.0.2rc3	cached
rocm-10.1-ga_rccl-develop-perf	rccl-develop	develop	`dc657f2`	cached
rocm-10.1-ga_rccl-develop-perf	rcclt-develop-perf	develop	`dc657f2`	cached

Failures

No failing tests in this run.

15 tests skipped.

aduforest and others added 4 commits October 20, 2026 11:05

test(rccl): place the NET_OPTIONAL_RECV_COMPLETION stub next to rcclU…

f921ea4

…seAinic Co-Authored-By: Claude <noreply@anthropic.com>

github-actions Bot added the project: rccl label Oct 20, 2026

systems-assistant Bot added the organization: ROCm label Oct 20, 2026

amd-argus Bot reviewed Oct 20, 2026

View reviewed changes

Merge origin/develop into users/aduforest/aicomrccl-2813-net-ib-cast-…

b0cf375

…init-host-microtests # Conflicts: # projects/rccl/tools/scripts/test_runner/lib/test_executor.py

amd-argus Bot reviewed Oct 21, 2026

View reviewed changes

aduforest marked this pull request as ready for review October 21, 2026 00:34

aduforest requested review from a team October 21, 2026 00:34

test(rccl): host-only suites declare num_gpus 0

f66e1e4

Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus Bot reviewed Oct 21, 2026

View reviewed changes

Comment thread projects/rccl/test/host/net-ib-cast-init-test.cc Outdated

Comment thread projects/rccl/test/host/net-ib-cast-gin-test.cc

test(rccl): skip leaking gin error-path tests under ASan; make the IB…

5b15c49

…_DISABLE test bite Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus Bot reviewed Oct 21, 2026

View reviewed changes

Comment thread projects/rccl/test/host/net-ib-cast-init-test.cc

Comment thread projects/rccl/test/host/net-ib-cast-init-test.cc

test(rccl): make the no-verbs init test bite; give merged devices rea…

a7e2c2e

…l PCI roots Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus Bot reviewed Oct 21, 2026

View reviewed changes

aduforest and others added 2 commits October 21, 2026 20:50

Merge remote-tracking branch 'origin/develop' into users/aduforest/ai…

79e7fdd

…comrccl-2813-net-ib-cast-init-host-microtests

amd-argus Bot reviewed Oct 22, 2026

View reviewed changes

Merge remote-tracking branch 'origin/develop' into users/aduforest/ai…

3d9a2d4

…comrccl-2813-net-ib-cast-init-host-microtests

Merge remote-tracking branch 'origin/develop' into users/aduforest/ai…

8ef1ac9

…comrccl-2813-net-ib-cast-init-host-microtests

amd-argus Bot reviewed Oct 22, 2026

View reviewed changes

Comment thread projects/rccl/test/host/net-ib-cast-init-test.cc

aduforest and others added 2 commits October 22, 2026 14:18

test(rccl): address ARGUS findings on net_ib_cast init tests

dbe4f99

Co-Authored-By: Claude <noreply@anthropic.com>

Merge remote-tracking branch 'origin/develop' into users/aduforest/ai…

48bcb85

…comrccl-2813-net-ib-cast-init-host-microtests

amd-argus Bot reviewed Oct 22, 2026

View reviewed changes

test(rccl): name the double-free pin's ticket

305951a

Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus Bot reviewed Oct 22, 2026

View reviewed changes

pvallem reviewed Oct 24, 2026

View reviewed changes

Comment thread projects/rccl/test/host/net-ib-cast-init-test.cc

aduforest and others added 2 commits October 23, 2026 17:40

Merge remote-tracking branch 'origin/develop' into users/aduforest/ai…

e674fbd

…comrccl-2813-net-ib-cast-init-host-microtests

test(rccl): set up device 3 so only the range check rejects it per re…

bba9fe1

…view Co-Authored-By: Claude <noreply@anthropic.com>

aduforest requested a review from a team as a code owner October 24, 2026 01:17

test(rccl): leave ci-precheckin.json to its code owners

62483ad

Host CI runs these binaries through run_host_tests.sh, so the json entries are not needed. Co-Authored-By: Claude <noreply@anthropic.com>

amd-argus Bot reviewed Oct 24, 2026

View reviewed changes

Conversation

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

Uh oh!

What

How

Coverage

Mutation Testing

Test Plan

Uh oh!

aduforest commented Oct 20, 2026

Uh oh!

therock-pr-bot Bot commented Oct 20, 2026

✅ All Policy Checks Passed

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Findings

Uh oh!

aduforest commented Oct 20, 2026

Uh oh!

aduforest commented Oct 20, 2026

Uh oh!

aduforest commented Oct 20, 2026

Uh oh!

aduforest commented Oct 20, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aduforest commented Oct 21, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

aduforest commented Oct 21, 2026

Uh oh!

aduforest commented Oct 21, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

aduforest commented Oct 21, 2026

Uh oh!

aduforest commented Oct 21, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aduforest commented Oct 22, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aduforest commented Oct 22, 2026

Uh oh!

aduforest commented Oct 22, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

aduforest commented Oct 22, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aduforest commented Oct 22, 2026

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

amd-argus Bot left a comment

Choose a reason for hiding this comment

Uh oh!

aduforest commented Oct 24, 2026

aduforest commented Oct 20, 2026 •

edited

Loading