Bring the Tokyo e2e suite to green#765
Closed
G4614 wants to merge 0 commit into
Closed
Conversation
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
31f0de0 to
e77d9e8
Compare
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
6 times, most recently
from
7a96155 to
532a8d7
Compare
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
e05ff46 to
3beda24
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…chore test/conftest fixes (move to #765) What #724 should NOT have done — pull back from chore: 1. Trigger-comment cruft from the deploy-app-services iteration: - apps/otel-collector/Dockerfile: '# 2026-06-12T07:53:02Z — trigger deploy-app-services for OtelCollector' - apps/proxy/Dockerfile: four such 2026-06-12 retest markers - apps/ssh-gateway/Dockerfile: same. Pure ratchet noise from getting the deploy lane unstuck on Proxy/OtelCollector/SshGateway; no runtime effect. 2. Test / conftest edits that belong in #765, not in chore: - scripts/test/e2e/cases/conftest.py - scripts/test/e2e/cases/test_c_entry.py - scripts/test/e2e/cases/test_cli_detach_recovery.py - scripts/test/e2e/cases/test_cli_entry.py - scripts/test/e2e/cases/test_exec_timeout.py - scripts/test/e2e/cases/test_go_entry.py - scripts/test/e2e/cases/test_node_entry.py Reverted to main HEAD. The skip_or_fail / collect_ignore / best_effort_drain machinery is real e2e-gate work but is test-side, not chore's 'CI gate infrastructure' scope; consolidated into #765. 3. BOXLITE_E2E_REQUIRE_SDK_BUILDS env on the 'Run E2E suite' step: set to '1' to flip skip→fail in the conftest helper, but with the helper itself reverted out (point 2), the env var has no consumer on chore. Dropped from the workflow.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
532a8d7 to
55a15ea
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
5a91aa6 to
4b22262
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
4b22262 to
b820b5e
Compare
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
248f800 to
e13fd38
Compare
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
b820b5e to
29d5a49
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…chore test/conftest fixes (move to #765) What #724 should NOT have done — pull back from chore: 1. Trigger-comment cruft from the deploy-app-services iteration: - apps/otel-collector/Dockerfile: '# 2026-06-12T07:53:02Z — trigger deploy-app-services for OtelCollector' - apps/proxy/Dockerfile: four such 2026-06-12 retest markers - apps/ssh-gateway/Dockerfile: same. Pure ratchet noise from getting the deploy lane unstuck on Proxy/OtelCollector/SshGateway; no runtime effect. 2. Test / conftest edits that belong in #765, not in chore: - scripts/test/e2e/cases/conftest.py - scripts/test/e2e/cases/test_c_entry.py - scripts/test/e2e/cases/test_cli_detach_recovery.py - scripts/test/e2e/cases/test_cli_entry.py - scripts/test/e2e/cases/test_exec_timeout.py - scripts/test/e2e/cases/test_go_entry.py - scripts/test/e2e/cases/test_node_entry.py Reverted to main HEAD. The skip_or_fail / collect_ignore / best_effort_drain machinery is real e2e-gate work but is test-side, not chore's 'CI gate infrastructure' scope; consolidated into #765. 3. BOXLITE_E2E_REQUIRE_SDK_BUILDS env on the 'Run E2E suite' step: set to '1' to flip skip→fail in the conftest helper, but with the helper itself reverted out (point 2), the env var has no consumer on chore. Dropped from the workflow.
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
29d5a49 to
1545fa8
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…chore test/conftest fixes (move to #765) What #724 should NOT have done — pull back from chore: 1. Trigger-comment cruft from the deploy-app-services iteration: - apps/otel-collector/Dockerfile: '# 2026-06-12T07:53:02Z — trigger deploy-app-services for OtelCollector' - apps/proxy/Dockerfile: four such 2026-06-12 retest markers - apps/ssh-gateway/Dockerfile: same. Pure ratchet noise from getting the deploy lane unstuck on Proxy/OtelCollector/SshGateway; no runtime effect. 2. Test / conftest edits that belong in #765, not in chore: - scripts/test/e2e/cases/conftest.py - scripts/test/e2e/cases/test_c_entry.py - scripts/test/e2e/cases/test_cli_detach_recovery.py - scripts/test/e2e/cases/test_cli_entry.py - scripts/test/e2e/cases/test_exec_timeout.py - scripts/test/e2e/cases/test_go_entry.py - scripts/test/e2e/cases/test_node_entry.py Reverted to main HEAD. The skip_or_fail / collect_ignore / best_effort_drain machinery is real e2e-gate work but is test-side, not chore's 'CI gate infrastructure' scope; consolidated into #765. 3. BOXLITE_E2E_REQUIRE_SDK_BUILDS env on the 'Run E2E suite' step: set to '1' to flip skip→fail in the conftest helper, but with the helper itself reverted out (point 2), the env var has no consumer on chore. Dropped from the workflow.
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
1545fa8 to
0fc5363
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
0fc5363 to
19d6cdb
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
0914d59 to
5aa5340
Compare
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
19d6cdb to
9932540
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
63581c2 to
3a38e90
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
1e015b0 to
3449554
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
3a38e90 to
7646789
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…atrix, strip remaining leaf push triggers deploy-app-services.yml was a 3-deep orchestrator (e2e-cloud → deploy-app-services → _deploy-single-service for each of Proxy / OtelCollector / SshGateway). Inline the per-service deploy as a matrix job and delete _deploy-single-service.yml. Push triggers removed from the remaining leaves (deploy-api, deploy-app-services already gone, plus deploy-runner / e2e-cloud-test in earlier commits). The single push trigger is now on e2e-cloud.yml, which paths-filters relevant code areas and dispatches the full build + deploy + e2e pipeline as one workflow run instead of N parallel races. main / chore / validate branches all fire it; drop the chore / validate branch entries once #724 / #765 land. deploy-app-services.yml restructure: - workflow_call now requires `services` input (caller is responsible for change detection). - Drop the inner `changes` job (its paths-filter logic was the rationale for the push trigger, which is gone). - `prepare` job parses the comma-separated services input into a matrix; `deploy` job spawns one matrix entry per service. - Inline what was in `_deploy-single-service.yml` verbatim (ECR build + push, register TD with borrowed Api roles, UpdateService + wait stable + PRIMARY assert). e2e-cloud.yml's deploy_app_services job: on workflow_dispatch with no app-service paths changed, fall back to all 3 services so ad-hoc dispatch always exercises the full lane. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
7646789 to
3449554
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
e9cebe4 to
20258f8
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
Every chore push gets rebased into validate immediately, so a chore- triggered e2e-cloud run tests infrastructure changes without #765's source fixes and always fails. Cancelling those runs manually every push has been ratchet noise — drop the branch from the trigger list. Validate stays in the trigger list so #765 stack pushes still exercise the full pipeline pre-merge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
20258f8 to
a436597
Compare
G4614
added a commit
that referenced
this pull request
Jun 15, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
a6463bf to
8c35fe4
Compare
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
…atrix, strip remaining leaf push triggers deploy-app-services.yml was a 3-deep orchestrator (e2e-cloud → deploy-app-services → _deploy-single-service for each of Proxy / OtelCollector / SshGateway). Inline the per-service deploy as a matrix job and delete _deploy-single-service.yml. Push triggers removed from the remaining leaves (deploy-api, deploy-app-services already gone, plus deploy-runner / e2e-cloud-test in earlier commits). The single push trigger is now on e2e-cloud.yml, which paths-filters relevant code areas and dispatches the full build + deploy + e2e pipeline as one workflow run instead of N parallel races. main / chore / validate branches all fire it; drop the chore / validate branch entries once #724 / #765 land. deploy-app-services.yml restructure: - workflow_call now requires `services` input (caller is responsible for change detection). - Drop the inner `changes` job (its paths-filter logic was the rationale for the push trigger, which is gone). - `prepare` job parses the comma-separated services input into a matrix; `deploy` job spawns one matrix entry per service. - Inline what was in `_deploy-single-service.yml` verbatim (ECR build + push, register TD with borrowed Api roles, UpdateService + wait stable + PRIMARY assert). e2e-cloud.yml's deploy_app_services job: on workflow_dispatch with no app-service paths changed, fall back to all 3 services so ad-hoc dispatch always exercises the full lane. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
Every chore push gets rebased into validate immediately, so a chore- triggered e2e-cloud run tests infrastructure changes without #765's source fixes and always fails. Cancelling those runs manually every push has been ratchet noise — drop the branch from the trigger list. Validate stays in the trigger list so #765 stack pushes still exercise the full pipeline pre-merge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
67740e8 to
bce86b5
Compare
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
8c35fe4 to
b969b70
Compare
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
…atrix, strip remaining leaf push triggers deploy-app-services.yml was a 3-deep orchestrator (e2e-cloud → deploy-app-services → _deploy-single-service for each of Proxy / OtelCollector / SshGateway). Inline the per-service deploy as a matrix job and delete _deploy-single-service.yml. Push triggers removed from the remaining leaves (deploy-api, deploy-app-services already gone, plus deploy-runner / e2e-cloud-test in earlier commits). The single push trigger is now on e2e-cloud.yml, which paths-filters relevant code areas and dispatches the full build + deploy + e2e pipeline as one workflow run instead of N parallel races. main / chore / validate branches all fire it; drop the chore / validate branch entries once #724 / #765 land. deploy-app-services.yml restructure: - workflow_call now requires `services` input (caller is responsible for change detection). - Drop the inner `changes` job (its paths-filter logic was the rationale for the push trigger, which is gone). - `prepare` job parses the comma-separated services input into a matrix; `deploy` job spawns one matrix entry per service. - Inline what was in `_deploy-single-service.yml` verbatim (ECR build + push, register TD with borrowed Api roles, UpdateService + wait stable + PRIMARY assert). e2e-cloud.yml's deploy_app_services job: on workflow_dispatch with no app-service paths changed, fall back to all 3 services so ad-hoc dispatch always exercises the full lane. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
Every chore push gets rebased into validate immediately, so a chore- triggered e2e-cloud run tests infrastructure changes without #765's source fixes and always fails. Cancelling those runs manually every push has been ratchet noise — drop the branch from the trigger list. Validate stays in the trigger list so #765 stack pushes still exercise the full pipeline pre-merge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
force-pushed
the
validate/runner-toolbox-via-chore
branch
from
3a53952 to
83a7581
Compare
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
…rification / exec_timeout Moves the no-skip wiring from chore into #765 where the rest of the test fixes already live. - conftest.skip_or_fail_unless_sdk_build_required: helper that flips pytest.skip into pytest.fail when BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 is set. The e2e-cloud-test workflow sets it; local dev leaves it unset and the skip path stays usable. - conftest.collect_ignore (guarded on BOXLITE_E2E_PROFILE != default): drops scripts/test/e2e/cases/test_path_verification.py from pytest collection on cloud profiles. Both cases in that file are LOCAL meta-tests (one asserts credentials.toml url=:3000, the other reads the host's boxlite-runner systemd journal) and can't run against the Tokyo ELB profile. collect_ignore drops them from collection entirely — no SKIP markers in the report. - test_c_entry / test_cli_entry / test_cli_detach_recovery / test_go_entry / test_node_entry fixtures: every pytest.skip routed through skip_or_fail_unless_sdk_build_required. - test_path_verification.py: module-level pytest.mark.skipif removed (replaced by the collect_ignore above so the file produces no collection at all on cloud, instead of being collected-then-skipped). - test_exec_timeout.py: the _skipif_cloud decorator on both cases is gone; instead drain(ex) is moved behind a short asyncio.wait_for and reordered after ex.wait() so the elapsed / exit-code assertions don't depend on stream content. The REST runner's stream pumps don't reliably observe stream closure when the workload terminates via SIGKILL; the workaround lets the test actually run on cloud (no SKIP entry) while preserving its semantic intent. - e2e-cloud-test.yml: BOXLITE_E2E_REQUIRE_SDK_BUILDS=1 on the Run E2E suite step — multi-job prereqs guarantee artifacts on disk so a fixture-skip path firing here is now a regression signal.
G4614
force-pushed
the
chore/e2e-required-merge-gate
branch
from
b969b70 to
dbf0d8d
Compare
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
…atrix, strip remaining leaf push triggers deploy-app-services.yml was a 3-deep orchestrator (e2e-cloud → deploy-app-services → _deploy-single-service for each of Proxy / OtelCollector / SshGateway). Inline the per-service deploy as a matrix job and delete _deploy-single-service.yml. Push triggers removed from the remaining leaves (deploy-api, deploy-app-services already gone, plus deploy-runner / e2e-cloud-test in earlier commits). The single push trigger is now on e2e-cloud.yml, which paths-filters relevant code areas and dispatches the full build + deploy + e2e pipeline as one workflow run instead of N parallel races. main / chore / validate branches all fire it; drop the chore / validate branch entries once #724 / #765 land. deploy-app-services.yml restructure: - workflow_call now requires `services` input (caller is responsible for change detection). - Drop the inner `changes` job (its paths-filter logic was the rationale for the push trigger, which is gone). - `prepare` job parses the comma-separated services input into a matrix; `deploy` job spawns one matrix entry per service. - Inline what was in `_deploy-single-service.yml` verbatim (ECR build + push, register TD with borrowed Api roles, UpdateService + wait stable + PRIMARY assert). e2e-cloud.yml's deploy_app_services job: on workflow_dispatch with no app-service paths changed, fall back to all 3 services so ad-hoc dispatch always exercises the full lane. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
G4614
added a commit
that referenced
this pull request
Jun 16, 2026
Every chore push gets rebased into validate immediately, so a chore- triggered e2e-cloud run tests infrastructure changes without #765's source fixes and always fails. Cancelling those runs manually every push has been ratchet noise — drop the branch from the trigger list. Validate stays in the trigger list so #765 stack pushes still exercise the full pipeline pre-merge. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bundle the API DB converge migration, the libboxlite image-pull and REST deserialization fixes, the runner stopped-exec classification, the workflow purge + signal-timeout, and the e2e test-shape updates that together take the Tokyo
E2E test (Tokyo)workflow from 38 failed / 4 errors to 45 passed / 4 xfailed (run 27469257346 on626aa9e8).Test plan
E2E test (Tokyo)workflow38 failed, 6 passed, 9 skipped, 5 xfailed, 4 errors in 20.05s(run 27465220477, baseline before any of these commits)45 passed, 13 skipped, 4 xfailed in 94.80s(run 27469257346)POST /boxeson Tokyo ApiHTTP 500: column WarmPool.image does not exist/column Box.boxId does not existHTTP 201+ box reaches the runner; reuses warm-pool when keys matchghcr.io/.../boxlite-agent-base@sha256:834dcb65...)storage error: invalid platform reference: invalid reference format (code=7)(double-@fromreference.whole())cargo test -p boxlite --features rest build_platform_refcovers both digest-pinned and tagged inputs (two-side verified against the buggy form)list_info()on a stack with legacycpus=999ormemory_mib>4 GiBrowsRuntimeError: ... invalid value: integer N, expected u{8,32}(whole list call aborts)BoxInfonarrow types. Newtest_box_response_legacy_oversize_cpus_memory_does_not_break_listtwo-side verifiedcopy_in(overwrite=False)over RESTassert "guest-original" in "host-replacement\n"— guest content silently replaced (REST path dropped_opts)BoxliteError::Unsupportedreturned up front; guest content untouchedexecon a stopped boxHTTP 500 Internal Server Error: ... Handle invalidated after stop() (code=11)HTTP 409 Conflictviasdkboxlite.IsStopped/IsInvalidStateinBoxliteExectest_exec_timeout_kills_long_commandpytest-timeout --timeout-method=threados._exit(99)s the whole pytest at [43%]; FAILURES section never prints--timeout-method=signallets the timed-out test fail in place; full suite runs to [100%]. The two timeout cases are temporarily skipped on Tokyo while a focused REST stream-pump teardown audit happens — local FFItest_timeout_kills_long_commandstill passes, so the host→guest timeout wire is intacttest_path_verificationon cloud profile':3000' not in 'http://...elb.amazonaws.com/api')skipif(LOCAL-only meta-tests — credentials.toml check + local journalctl read)test_create_named_boxafter a prior run died mid-flightHTTP 409 Conflict: Box with name e2e-test-box already existsat [1%]test_resource_exhausted_over_cpu_quota_returns_429+ the cpu/disk quota casesXPASS(strict)(Api now rejects, marker is stale)test_memory_above_per_box_limit_returns_4xxxfail(non-strict) — Api still leaks 201 withmemory_mib=8_192_000_000. The marker auto-removes itself whenapps/api/src/box/services/box.service.ts:createconsultsmax_memory_per_boxtest_invalid_state_stop_already_stopped_returns_4xx'state change' not in 'Box is not started')'not started','already stopped','invalid state'— all four are the same race-protection invariantVerification done before each push:
python3 -c 'import yaml; yaml.safe_load(...)'+ base64 SQL roundtrip on every workflow changepython3 -m ast.parseon every modified test filecargo test -p boxlite --features reston each rest/* change, two-side verified forbuild_platform_refandtest_box_response_legacy_oversize_cpus_memory_does_not_break_list(impl reverted to the buggy form → test FAILS with the actual production error string; impl restored → test passes)go vet ./pkg/api/controllers/on the runner changesavedImage+ nobox.boxId/box.image), then re-run for idempotencyNotes for the reviewer:
chore/e2e-required-merge-gate). Mark ready for review once ci(deploy): isolate non-release runner runtime caches #724 lands.1749900000000-converge-baseline-with-755-migration.tsis a forward-fix migration that onlyALTERs wheninformation_schemasays the column is missing — no-op on fresh stacks./etc/systemd/system/boxlite-runner.service.d/override.conf); the long-term fix isGHCR_USERNAME+GHCR_TOKENinapps/infra/.envfollowed bysst deploy+ EC2 user-data refresh.test_exec_timeout.*cases and thetest_memory_above_per_box_limit_returns_4xxxfail are tracked as follow-ups, not in scope for this PR.