Repository navigation
Description
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
- CI — https://github.com/softmata/horus/actions/runs/32628222536
- Cross-Platform Parity — https://github.com/softmata/horus/actions/runs/32628222533
- Safety — https://github.com/softmata/horus/actions/runs/32614104941
Recurring unexplained failures — these need attention
Integration Tests — failed 2 times
- 2026-09-02T21:09:12Z 3246dd6 https://github.com/softmata/horus/actions/runs/33683509747
- 2026-08-23T08:25:26Z 8080ae4 https://github.com/softmata/horus/actions/runs/32628222567
log tail
2026-09-02T22:13:21.6507480Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:21.6636031Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:21.6764480Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:21.6894010Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:21.7022552Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:21.7159254Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:22.7297490Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-02T22:13:38.0814144Z [multithread] allocs: 20931647, failures: 0, releases: 20931641, corruptions: 0
2026-09-02T22:15:51.0106105Z thread 'test_slot_reuse_after_release' (104204) panicked at horus_core/tests/tensor_pool_concurrent.rs:145:62:
2026-09-02T22:15:51.0118950Z test test_slot_reuse_after_release ... FAILED
2026-09-02T22:15:51.0119578Z failures:
2026-09-02T22:15:51.0119884Z failures:
2026-09-02T22:15:51.0120986Z test result: FAILED. 4 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.14s
2026-09-02T22:15:51.0159639Z ##[error]Process completed with exit code 101.
Multi-Platform — failed 6 times
- 2026-09-01T20:41:58Z 3765e3f https://github.com/softmata/horus/actions/runs/33556838275
- 2026-08-31T10:11:33Z 1e48195 https://github.com/softmata/horus/actions/runs/33381276097
- 2026-08-31T05:08:06Z 42d66ed https://github.com/softmata/horus/actions/runs/33359433636
log tail
2026-09-01T21:21:43.1082262Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-01T21:21:43.1929780Z [ERROR] [unknown] Fatal init: Node error: Node 'panic_init_node' panicked during init
2026-09-01T21:21:43.1985939Z Failed to initialize node 'panic_init_node': Node error: Node 'panic_init_node' panicked during init
2026-09-01T21:21:43.1986835Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-01T21:21:43.1991250Z [ERROR] [unknown] Fatal init: Node error: Node 'panic_init_node' panicked during init
2026-09-01T21:21:43.1991532Z Failed to initialize node 'panic_init_node': Node error: Node 'panic_init_node' panicked during init
2026-09-01T21:21:43.3214472Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-01T21:21:43.3813284Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-01T21:21:43.5842481Z Warning: Failed to set signal handler: Ctrl-C error: Ctrl-C signal handler already registered
2026-09-01T21:21:44.3281997Z failures:
2026-09-01T21:21:44.3282673Z thread 'communication::topic::tests::broadcast_subscriber_join_midstream' (9763) panicked at horus_core/src/communication/topic/tests.rs:7163:5:
2026-09-01T21:21:44.3285571Z failures:
2026-09-01T21:21:44.3286083Z test result: FAILED. 2460 passed; 1 failed; 4 ignored; 0 measured; 4 filtered out; finished in 72.72s
2026-09-01T21:22:23.5884978Z ##[error]Process completed with exit code 101.
Docs Contract — failed 11 times
- 2026-09-01T09:10:04Z d5b7720 https://github.com/softmata/horus/actions/runs/33490785991
- 2026-08-31T10:31:53Z 1e48195 https://github.com/softmata/horus/actions/runs/33382822599
- 2026-08-30T09:43:57Z c9fa8be https://github.com/softmata/horus/actions/runs/33304764871
log tail
2026-09-01T09:15:07.2539488Z content/docs/troubleshooting.mdx:939 src/lib.rs:298:29: error[E0609]: no field 'topic' on type '&mut docblock_578::_DocSelf': unknown field
2026-09-01T09:15:07.2540508Z content/docs/troubleshooting.mdx:950 src/lib.rs:313:20: error[E0609]: no field 'topic' on type '&mut docblock_579::_DocSelf': unknown field
2026-09-01T09:15:07.2541511Z content/docs/troubleshooting.mdx:971 src/lib.rs:344:13: error[E0609]: no field 'sensor' on type '&mut docblock_581::_DocSelf': unknown field
2026-09-01T09:15:07.2542509Z content/docs/troubleshooting.mdx:982 src/lib.rs:360:21: error[E0609]: no field 'sensor' on type '&mut docblock_582::_DocSelf': unknown field
2026-09-01T09:15:07.2543584Z content/docs/troubleshooting.mdx:992 src/lib.rs:373:5: error[E0433]: cannot find module or crate 'thread' in this scope: use of unresolved module or unlinked crate 'thread'
2026-09-01T09:15:07.2544812Z content/docs/troubleshooting.mdx:1052 src/lib.rs:386:37: error[E0424]: expected value, found module 'self': 'self' value is a keyword only available in methods with a 'self' parameter
2026-09-01T09:15:07.2548898Z thread 'documented_rust_examples_compile' (9229) panicked at horus_manager/tests/docs_examples.rs:1241:9:
2026-09-01T09:15:07.2550430Z src/lib.rs:203:16: error[E0061]: this function takes 4 arguments but 3 arguments were supplied
2026-09-01T09:15:07.2551475Z src/lib.rs:203:16: error[E0277]: the '?' operator can only be applied to values that implement 'Try': the '?' operator cannot be applied to type 'horus::types::OccupancyGrid'
2026-09-01T09:15:07.2561017Z test documented_rust_examples_compile ... FAILED
2026-09-01T09:15:07.2561315Z failures:
2026-09-01T09:15:07.2561506Z failures:
2026-09-01T09:15:07.2562163Z test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 26 filtered out; finished in 44.80s
2026-09-01T09:15:07.2573183Z ##[error]Process completed with exit code 101.
Feature Matrix — failed 3 times
- 2026-08-24T03:46:26Z 8080ae4 https://github.com/softmata/horus/actions/runs/32687618386
- 2026-08-23T08:25:26Z 8080ae4 https://github.com/softmata/horus/actions/runs/32628222552
- 2026-08-23T00:58:46Z 59a516a https://github.com/softmata/horus/actions/runs/32609220131
log tail
2026-08-24T03:47:14.1689955Z error[E0559]: variant 'BlackBoxEvent::NodeError' has no field named 'severity'
2026-08-24T03:47:14.1691672Z 5030 | severity: crate::error::Severity::Fatal,
2026-08-24T03:47:14.1888024Z error[E0559]: variant 'BlackBoxEvent::NodeError' has no field named 'severity'
2026-08-24T03:47:14.1889341Z 592 | ... severity: crate::error::Severity::Fatal,
2026-08-24T03:47:14.2028550Z error[E0559]: variant 'BlackBoxEvent::NodeError' has no field named 'severity'
2026-08-24T03:47:14.2030352Z 389 | ... severity: crate::error::Severity::Fatal,
2026-08-24T03:47:14.2056749Z error[E0559]: variant 'BlackBoxEvent::NodeError' has no field named 'severity'
2026-08-24T03:47:14.2058187Z 280 | ... severity: crate::error::Severity::Fatal,
2026-08-24T03:47:14.2091563Z error[E0559]: variant 'BlackBoxEvent::NodeError' has no field named 'severity'
2026-08-24T03:47:14.2092779Z 344 | ... severity: crate::error::Severity::Fatal,
2026-08-24T03:47:15.1737143Z ##[error]Process completed with exit code 101.
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
- Timing, throughput and tick-rate assertions are only meaningful with
--release. In debug they fail for no reason. --test-threads=1is required; the tests share one SHM namespace.- Run
rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelated
tests fail in a rotating pattern. - Without
--no-fail-fast, cargo stops at the first failing binary and
you see one failure per run. - A bare
exit code 1with no output is usually the/tmpper-user
quota, not your change.dflooks fine because it is a quota. - Detection of the concurrency bugs here is probabilistic. Passing once
is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a loom model
(horus_core/tests/loom_*.rs). It explores exhaustively and models weak
memory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.md files to this repo; it is public. Do not add co-authorship
trailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
Activity
github-actions commentedon Sep 6, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 12 timeslog tail
Integration Tests— failed 2 timeslog tail
Multi-Platform— failed 6 timeslog tail
Feature Matrix— failed 2 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 7, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 12 timeslog tail
Integration Tests— failed 2 timeslog tail
Multi-Platform— failed 6 timeslog tail
Feature Matrix— failed 2 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 8, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 13 timeslog tail
Integration Tests— failed 2 timeslog tail
Multi-Platform— failed 5 timeslog tail
Feature Matrix— failed 2 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 9, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 13 timeslog tail
Integration Tests— failed 2 timeslog tail
Multi-Platform— failed 5 timeslog tail
Feature Matrix— failed 2 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 10, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 14 timeslog tail
Integration Tests— failed 2 timeslog tail
Multi-Platform— failed 5 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 11, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 15 timeslog tail
Multi-Platform— failed 5 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 12, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 16 timeslog tail
Multi-Platform— failed 5 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 13, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 16 timeslog tail
Multi-Platform— failed 5 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 14, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 15, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 16, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 17, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 18, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.
github-actions commentedon Sep 19, 2026
Autopilot triage — 25 failed run(s) on main
Unexplained, single occurrence (watching, not reporting)
Recurring unexplained failures — these need attention
Docs Contract— failed 17 timeslog tail
Multi-Platform— failed 4 timeslog tail
Before changing any code
Most red CI in this repo is environmental. The triage above already
filtered the failures it could explain, but confirm the remainder is
real before touching source:
--release. In debug they fail for no reason.--test-threads=1is required; the tests share one SHM namespace.rm -rf /dev/shm/horus_*first. Leaked namespaces make unrelatedtests fail in a rotating pattern.
--no-fail-fast, cargo stops at the first failing binary andyou see one failure per run.
exit code 1with no output is usually the/tmpper-userquota, not your change.
dflooks fine because it is a quota.is not evidence of a fix.
The rule that matters most
Do not claim a fix without an ablation: show the test FAILS without the
change and passes with it. A test that passes both ways proves nothing,
and that is the most common way work here goes wrong. If you cannot
construct a failing case, say so and leave the finding documented
rather than shipping a change you cannot demonstrate.
For anything about ordering or interleaving, prefer a
loommodel(
horus_core/tests/loom_*.rs). It explores exhaustively and models weakmemory ordering, which x86 cannot exhibit. Every loom model here
documents the mutation that turns it red — yours must too.
Scope
Fix only what this issue reports. If you find something else, say so
separately rather than widening the change. Do not add scratch or report
.mdfiles to this repo; it is public. Do not add co-authorshiptrailers. Explain the mechanism in the commit message, with the numbers
you actually measured.