Repository navigation
fix: the live Box Egress Proxy e2e cases run on the Linux KVM lane and no longer report NOT RUN for - #3690
Conversation
bep_live_gate no longer stops every Linux lane. On Linux it gates on a TPM or Secret Service that answers, or on E2E_BEP_KEY_STORE=1 from the lane. The Keychain-only store-reference case and the dns-steering store leg are NOT RUN on Linux with that reason, since min secret uses the macOS Keychain only. The Linux minvmd-build recipe now builds bep beside the musl minvmd, so neither minvmd nor rt_bep_bin needs MINVMD_BEP_BIN.
…ox-egress-proxy-e2e-cases-run-on-the-l
|
Important Review skippedBot user detected. Ask an organization administrator to verify that this bot has an active assigned CodeRabbit review seat. See the seat-assignment guide. After the seat is confirmed, request a review again.
|
|
Reviewed: no findings; approval withheld, 5 threads unresolved at APPROVE withheld: 5 unresolved threads — minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh:14196, minimal-arch[bot] scripts/session-e2e.sh:14186. Resolving the last thread releases the approval on this head. Resolve the open threads. Resolving the last one releases the approval on this head; no new review is required. Verdict APPROVE, posted as a comment: the App cannot review its own pull request; the check and the label carry it. Clobber: none seen. |
There was a problem hiding this comment.
arch-review: CONDITIONS · 3 conditions · the full ruling
The Linux gate and the recipe conform to Gatehouse §6.10. But the Linux NOT RUN legs that remain now have no issue that owns them, and the PR closes #3652. Section 6.10 line 844 and architecture.md line 553 leave the Linux min secret store open. File an issue for it, and name it in the store reasons (C1). No lane sets E2E_BEP_KEY_STORE, so the KVM lane does not yet show PASS. Change Fixes to Refs #3652, and name #3652 in the key-store reasons (C2). Remove the Darwin exit in proof_bep_ahm_missing_proxy_refused, because its reason is now false (C3).
Outside the diff
scripts/session-e2e.sh:19097: C3 · This reason says the proxy runs on macOS only, and this PR makes that false.check_github_grantsandproxy_absenceincrates/minimal/src/cmd/boxes/create.rshave no platform gate. Remove this Darwin block, and keep the gvproxy check.
Ruled at 04be33c against gominimal/arch@4f2eb87d5d12 by claude-opus-5-5.
|
/revise
Run |
|
Round 1 of 10 on a /revise note, pushed as
Verify Review threads the coder judged:
On Linux, no bep case reports NOT RUN because the Box Egress Proxy runs on macOS only. The missing-proxy case now runs on Linux when gvproxy is present. A Linux run without a key store prints a NOT RUN reason that names #3652. The Keychain-only store cases name #3695. The KVM lane does not show PASS for the bep cases yet, because no lane sets E2E_BEP_KEY_STORE. Notes from the agent: Revise round 1. C1: the bep_native_store_gate reason and the dns-steering store leg reason now end with '; Linux: #3695'. C2: both key-store NOT RUN reasons in bep_linux_key_store end with '; #3652'. C3: I removed the Darwin-only exit in proof_bep_ahm_missing_proxy_refused, and the gvproxy check stays. I cannot edit the PR body. The runner must change 'Fixes #3652' to 'Refs #3652' and list #3695 as a residual. This box has no KVM, no TPM, no Secret Service and no network to go.dev for gvproxy. So both test lines exit 0 with NOT RUN, not PASS, and the proof on the KVM lane is not observed here. The printed reasons contain issue URLs because the directive asks for them. |
There was a problem hiding this comment.
arch-review: CONDITIONS · 2 conditions · the full ruling
The script changes meet the earlier conditions. C3 is met, and the NOT RUN reasons now name #3652 and #3695. Two edits to the PR body are still open. C1: list the Linux min secret store as a residual with issue #3695 (architecture.md line 553, open gap 17). C2: change Fixes #3652 to Refs #3652. The Linux key-store NOT RUN lines name #3652 as their owner, and no lane shows the bep cases as PASS yet.
Ruled at d94b1ba against gominimal/arch@4f2eb87d5d12 by claude-opus-5-5.
There was a problem hiding this comment.
minimal-review — approval withheld
Verdict COMMENT — APPROVE withheld: 5 unresolved threads — minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh, minimal-arch[bot] scripts/session-e2e.sh:14196, minimal-arch[bot] scripts/session-e2e.sh:14186. No findings.
Panel: 3 of 3 personas ran (correctness, security, supply-chain).
|
/arch-review |
There was a problem hiding this comment.
arch-review: PASS_WITH_CONDITIONS · 1 condition · the full ruling
The change conforms to the design. On Linux, the gate now probes the TPM and Secret Service. Gatehouse §6.10 lines 837 and 844 name these stores for Linux. The min secret legs are NOT RUN on Linux and name #3695. That is correct, because architecture.md line 553 leaves the Linux default store open (gap 17). Nobody saw the task's proof yet: no Linux KVM lane prints PASS for the bep cases. C1 carries that proof, and it is filed as #3652, so the PR can merge with C1 open.
Ruled at d94b1ba against gominimal/arch@4f2eb87d5d12 by claude-opus-5-5.
|
/review |
Follow-up
Refs #3652
scripts/session-e2e.shjustfileTask marker
Proof
On Linux, bep_live_gate no longer prints the NOT RUN reason that the live cases run on macOS lanes only. A Linux lane with a TPM or Secret Service that answers, or with E2E_BEP_KEY_STORE=1, starts the proxy and runs the bep cases. If the proxy does not come up there, the case fails.
just minvmd-buildon Linux puts bep beside the musl minvmd. Only partly delivered: bep_store_reference_end_to_end cannot print PASS on Linux, becausemin secretuses the macOS Keychain only. It now prints NOT RUN with that reason.Residual: the Linux
min secretstore (architecture.md line 553, open gap 17). On Linux,bep::storereturnsStoreError::Unsupported, sobep_store_reference_end_to_endand the store leg ofbep_dns_steering_through_the_box_zone_resolverprint NOT RUN. Tracked in #3695.Residual: no lane sets
E2E_BEP_KEY_STOREor provides a TPM or Secret Service yet, so the Linux KVM lane does not show the bep cases as PASS. #3652 stays open for that.Notes from the agent: I did not see PASS on the KVM lane. This box has no /dev/kvm, no TPM, no secret-tool, no musl toolchain and no network to fetch gvproxy. So both test lines report NOT RUN here, with the new reasons. The Linux
cargo build -p bep --target <musl>is not built here. The release workflow builds bep for musl, so the target is known to work. The live gate on Linux uses a TPM (/dev/tpmrm0 that this user can read and write, plus tpm2_createprimary on PATH) or Secret Service (secret-tool lookup exits 0 or 1 with no error output), or the lane key E2E_BEP_KEY_STORE=1. The lane key skips the probe, so a proxy that then does not come up is a failure. A KVM lane must set it, or provide a store, to see PASS. The proof is not fully achievable for bep_store_reference_end_to_end: bep::store's non-macOS host module returns StoreError::Unsupported, somin secret setcannot work on Linux. That case and the store leg of bep_dns_steering_through_the_box_zone_resolver now record NOT RUN on Linux with that reason. The fail_closed leg of that case now checks the git box alone, or is NOT RUN when no leg started a box. A follow-up task is needed for a Linuxmin secretstore. Not changed: proof_bep_ahm_missing_proxy_refused still has its own Darwin check ('the Box Egress Proxy runs on macOS only'). It is outside bep_live_gate, and the task did not name it. It can probably run on Linux now. shellcheck reports no new findings. I did not run commitl…[truncated]Tests
bash scripts/session-e2e.sh bep_store_reference_end_to_endbash scripts/session-e2e.sh bep_second_box_needs_no_sign_inVerify
min task run agent-verifypassed ([tasks.verify]).Base
The base branch moved while the coder worked; it was merged in before the tests ran.