Repository navigation
T110 — min ls shows each box's address and egress posture, and states the naming surface as activate does - #2777
Conversation
The ListSessions reply gains a map keyed by box id. Each entry carries the host loopback address the box's ports answer at and its effective egress posture, resolved the way GetEffectiveSessionPolicy resolves it. Both fields are serde-defaulted, so an older daemon reads as unreported.
min ls gains an ADDRESS column after NAME, and --json carries each box's loopback_address. The EGRESS column prints the posture in the words min session policy uses, from one shared function, with the per-box enforcement note for an unenforced host-address box. The activate host line and the BOX NAMES row now state how a box name opens through one resolver function, so the two verbs use the same words in every arm.
|
Important Review skippedBot user detected. To trigger a single review, invoke the
|
|
Approved: every review thread resolved at @norrietaylor resolved the last thread, which releases the approval this review had withheld. Nothing to do. |
There was a problem hiding this comment.
arch-review: CONDITIONS · 3 conditions · the full ruling
The daemon, wire and listing code conform to §7.1 lines 259-264. The address and the posture come from the owners the design names. Three text surfaces still carry the old naming words. The KVM e2e at scripts/session-e2e.sh lines 9416 and 9424 will fail, and the PR says so. Fix it in C1. The spec on main does not contain NET-146 or NET-147, and its NET-142 requires the old min ls row. Land the amendment in C2. Update the CLI reference in C3.
Outside the diff
scripts/session-e2e.sh:9416: C1 · This pattern pins the old BOX NAMES words. After this change the row saysopens at its name on this machine. Change this pattern and the one at line 9424 to the new words. Otherwise the KVM lane fails.docs/specs/18-spec-box-networking/18-spec-box-networking.md:262: C2 · On main, this requirement saysmin lskeeps its old row wording, and this change rewrites the row. Mergeb15f1883first, or bring its NET-142, NET-146 and NET-147 hunks into this PR.docs/reference/cli-min.md:549: C3 · This quote is the old native clause. The next sentence says a browser opens the box only through the proxy. Replace both with the new behaviour: the box opens at its name, and the proxy opens it too.
Ruled at ac34f0a against gominimal/arch@8640f85e193d by claude-opus-5-5.
activate and the min ls BOX NAMES row now say a box name opens at its name on this machine, and name the HTTP proxy that opens it too. The native-daemon e2e still grepped for the old phrases.
The negative checks looked for the old phrases, so the new own-address wording would pass them. They now refuse any 'on this machine' claim.
There was a problem hiding this comment.
arch-review: CONDITIONS · 2 conditions · the full ruling
The listing work is in the right crates, and the address and posture follow §7.1 lines 259-261. Two items remain. C1: the spec on main still quotes the old NET-142 wording at lines 254 and 262, and NET-146 and NET-147 do not exist there. Land the text from b15f1883 first, or carry it in this PR. C2: boxes_at_own_address() treats every 127.0.0.1 as the fallback. Section 7.1 line 261 makes 127.0.0.1 a host-address box's real address on a native node. Report the fallback as a field from the daemon, and read the fallback form when any box is at the fallback.
Ruled at a151f39 against gominimal/arch@8640f85e193d by claude-opus-5-5.
|
Round 1 of 10 on a failed check suite, pushed as
Verify Outside the task's
Converted to a draft: something above needs a person before review. Review threads the coder judged:
Notes from the agent: Revise round 1. native-daemon-e2e failed because scripts/session-e2e.sh still grepped for the old wording: 'resolves on this machine' on the activate line, and 'resolves through this machine's DNS' and 'the hostname proxy also listens on 127.0.0.1:' on the BOX NAMES row. ci-linux-native-success is the aggregator and failed only because of that job. scripts/session-e2e.sh is OUTSIDE the task's layers. I changed it because the task changes text that it pins. The patterns now accept the own-address form and the 127.0.0.1-fallback form, and the BOX NAMES check looks for 'the HTTP proxy at 127.0.0.1:'. I did not run the e2e on this box. Its lane must prove it. In crates/minimal/tests/cli.rs, two negative checks looked for the old phrases. The new own-address wording passed them without effect, so they now refuse any 'on this machine' claim. Spec drift: NET-142's native-DNS sub-requirement still quotes 'names: .min.internal resolves on this machine; open it through the HTTP proxy at 127.0.0.1:' for every native host. The task fixes the own-address wording 'opens at its name on this machine; the HTTP proxy at 127.0.0.1: opens it too', and the code follows the task. The spec quote must change in follow-up work. |
|
/revise
|
The daemon now marks a listed box at_fallback when it stands on the 127.0.0.1 fallback. A host-address box at the node address is not at the fallback. min ls reads the fallback form when any box reports it, so a mixed listing no longer claims every name opens at its name.
grep -q under pipefail can SIGPIPE min ls and read as a gone session. On a miss the check now prints the listing, its exit status and stderr, and the attach transcript, so the next failure names its cause.
|
Round 2 of 10 on a /revise note, pushed as
Verify Outside the task's
Converted to a draft: something above needs a person before review. Review threads the coder judged:
Notes from the agent: native-daemon-e2e: I could not run the hooks case here, and I did not find a proven root cause. This host has no /sys and no cgroup filesystem, and the sandbox's read-only remount of /etc fails with EPERM, so every box exits at launch ( |
norrietaylor
left a comment
There was a problem hiding this comment.
Review of head edefaf7 against T110's plan block (#2696).
Blocking
- The
min lsaddress assertion the epic owner carried here from #2749 (comment on #2696) is not restored.fresh_install_range_lands_on_macosinscripts/session-e2e.shstill ends after the two HTTP checks: nomnl ls | grep -qF -- "$frm_addr", no header bullet, and the OK line does not mentionmin ls. Amin lsthat prints-or127.0.0.1for a box the zone answers at 127.0.64.x keeps the macOS lane green. crates/minimal/src/resolver.rs:5048callsbox_name_opens(&name, surface, proxy_port, true)withat_own_addresshard-coded, so the activate host line never prints the fallback form. With native DNS live and a box on the127.0.0.1interim, activate printsweb.min.internal opens at its name on this machine; …whilemin ls(which reads each box's report) prints<box>.min.internal resolves on this machine; open it through …: the drift the task's proof forbids.ls_name_surface_matches_activate_host_linecompares only thetruearm; make it cover both.
Non-blocking
docs/reference/cli-min.md:44-66(session list) does not document the ADDRESS column, the EGRESS posture words, or the--jsonkeysloopback_addressandegress_posture.scripts/session-e2e.sh:7833,:9330,:9427-9428accept either wording, so a native-range box regressing to the fallback form passes. Pin the form each state requires.- The posture words live in two places:
egress_posture_words(crates/minimal/src/cmd/boxes/policy.rs:83) formin session policyandmin ls, andEffectiveEgress::summary_label()(crates/boxes/src/lib.rs:518) formin dash(dash/render.rs:532). Route the dashboard through the same function.
Checked: the listing reply's serde defaults and old-reply test, box_network_facts posture through the same path as min session policy, the ADDRESS column and --json placement for single- and multi-VM listings, every posture mode, the round-2 harness fix (legitimate: capture then grep -F under pipefail), and the new user-facing strings.
|
/revise Fix the blocking findings in the review on edefaf7, and findings 3–5 as they fall inside the layers:
Non-blocking
|
|
That Nothing to type again: the command gets a rocket when its round starts, or a refusal that names the reason. No round was counted for it yet. |
There was a problem hiding this comment.
arch-review: CONDITIONS · 2 conditions · the full ruling
The change now reads the fallback form when any listed box is at the fallback. That fixes the mixed-listing claim. One code gap remains. In a microVM, box_network_facts reports 127.0.0.1 for a host-address box before the node's address is granted. Section 7.1 line 263 says such a box is never mirrored onto 127.0.0.1. Make the change in C1. The spec at NET-142 (lines 254 and 262) and docs/reference/cli-min.md still state the old wording. Amend them as C2 says.
Ruled at edefaf7 against gominimal/arch@8640f85e193d by claude-opus-5-5.
|
/revise Arch review on edefaf7 (CONDITIONS), C2, for the same round as the review above:
C1 (the spec's NET-142 text) waits on a decision about #2679; do not change the spec in this round. |
…ws-each-box-s-address-and-egress-postu
The BOX NAMES row reads the own-address form only when a box reports it is off the fallback. A withheld host-address box reads as the fallback, and the activate line follows the interim flag for every network mode.
…ws-each-box-s-address-and-egress-postu
Spec lint
Advisory. None of this blocks the merge; the review decides. Advisory, and never a merge gate. Linter: |
|
Round 4 of 10 on a /revise note, pushed as
Verify Outside the task's
Converted to a draft: something above needs a person before review. Review threads the coder judged:
min ls shows each box's own address in the ADDRESS column, and the EGRESS column uses the same posture words as min session policy. min ls --json carries loopback_address. The BOX NAMES row and the activate host line use the same words. Both read the own-address form only when the box is proved to be off the 127.0.0.1 fallback. In every other case, both read the proxy form. Notes from the agent: Revise round 4. Directive items: (1) The min ls address assertion in fresh_install_range_lands_on_macos was already on the branch (commit 9c45075). It uses capture-then-match on the box's row, not a bare pipe into grep -q, because an early grep exit can SIGPIPE the CLI under pipefail. (2) Already on the branch in fa465cb. This round also drops the HostNet override, so the activate line follows interim_loopback for every network mode. (3) NET-142 and its sub-requirements now carry the b15f188 text, and NET-146 and NET-147 are added. No other hunk is in the change. I was not able to run just lint-prose because vale is not installed on this box. (4) scripts/session-e2e.sh now pins the own-address form at the range-step check, at the second box's activate line, and at the BOX NAMES row. Each of these cases has the range present and boxes at their own leased addresses. (5) Out of scope: crates/minimal/src/dash/render.rs is not in the layers, so min dash still has its own posture words and does not use egress_posture_words. Scope: create.rs and session-e2e.sh are outside the layers. Earlier rounds touched them, and the directive items 2 and 4 require them. docs/reference/cli-min.md is restored to main, because T125 (#2796) owns it, so the branch no longer changes it. I did not identify the failed ci-linux-native-success check, because the log excerpt shows no cause. The listing tests and the whole minimal suite pass here. The scripts/session-e2e.sh changes were checked with bash -n only. The e2e lanes did not run on this box. |
|
Round 5 of 10 on a /revise note: the coder changed nothing. noop: this round's note is against head 0a46a2c. The commits after that head (834a370, f17fa99, b9136ef, 0c69610) already address items 1-4, and I found nothing more to change. (1) fresh_install_range_lands_on_macos in scripts/session-e2e.sh runs |
norrietaylor
left a comment
There was a problem hiding this comment.
Babysitter review of a66544c against T110 (#2696). Each item of the last /revise is closed:
- The
min lsaddress assertion is back infresh_install_range_lands_on_macos. - The activate host line follows the interim flag for every network mode. A withheld host-address box reads as the fallback (
box_at_fallback), andmin lsreads the own-address form only when a box proves it. - The spec carries only the NET-142, NET-146 and NET-147 hunks.
- The e2e pins the wording that each state requires.
minimald-root-integrationis green on this head.
Every lane aggregator is green. The PR no longer changes docs/reference/cli-min.md, so T125 (#2796) still owns that page.
There was a problem hiding this comment.
arch-review: PASS · the full ruling
The change conforms to NET design §7.1. Each box reports the address its ports answer at, and a VM node's host-address box reports none until the node gets its address, as line 263 says. The BOX NAMES row and the activate line now read the own-address form only when a box proves it. Otherwise they read the proxy form. The spec amendment and the stricter e2e checks are in this head. Separately, docs/reference/cli-min.md still quotes the old host line. File an issue for it, because the PR names T125 but links no issue.
Ruled at a66544c against gominimal/arch@8640f85e193d by claude-opus-5-5.
…ws-each-box-s-address-and-egress-postu
There was a problem hiding this comment.
arch-review: PASS · the full ruling
This push merges main and corrects one doc comment. The merge has no conflict resolution, and its changes do not touch the listing or the host line. The new comment on name_surface_line_at now matches boxes_at_own_address. A listing that does not prove an own address reads the fallback form, as §7.1 line 259 requires. The PASS on a66544c8 stands for this head.
Ruled at 11d3691 against gominimal/arch@8640f85e193d by claude-opus-5-5.
There was a problem hiding this comment.
minimal-review — approval withheld
Verdict COMMENT — APPROVE withheld: 11 unresolved threads — minimal-arch[bot] crates/minimal/src/resolver.rs:5080, minimal-arch[bot] crates/minimal/src/cmd/list.rs:950, minimal-arch[bot] crates/minimald-rpc/src/lib.rs:369, minimal-arch[bot] crates/minimald/src/rpc.rs:284, minimal-arch[bot] crates/minimal/src/resolver.rs:5080, …. No findings.
Panel: 3 of 3 personas ran (correctness, security, supply-chain).
|
Round 6 of 10 on the runtime's review findings, pushed as
Verify Outside the task's
Converted to a draft: something above needs a person before review. Review threads the coder judged:
Notes from the agent: Revise round 6 changes only the doc comment on |
There was a problem hiding this comment.
minimal-review — approved
Verdict APPROVE — every review thread is resolved; @norrietaylor closed the last one. The review of 11d3691f1363 withheld this approval on the threads open then. Posted as a comment: the App cannot approve its own pull request; the check and the label carry it.
…2777) The ListSessions reply gains a map keyed by box id with the host loopback address the box's ports answer at, its effective egress posture, and whether it stands on the 127.0.0.1 fallback. All three are serde-defaulted, so an older daemon reads as unreported. min ls gains an ADDRESS column and --json carries the same facts. The EGRESS column prints the posture in the words min session policy uses, from one shared function, and the activate host line and the BOX NAMES row state how a box name opens through one resolver function. A VM node's host-address box reports no address until the node is granted one, since it is never mirrored onto 127.0.0.1. Backport of #2777. Adapted to the line: the policy.rs and create.rs hunks land in cmd/session.rs, and the minimald-rpc and minimald rpc tests land inline in lib.rs and rpc.rs, with the line's names (SessionStatus, EffectiveSessionPolicy, a built CreateSessionRequest). (cherry picked from commit 08045bf)
…2777) The ListSessions reply gains a map keyed by box id with the host loopback address the box's ports answer at, its effective egress posture, and whether it stands on the 127.0.0.1 fallback. All three are serde-defaulted, so an older daemon reads as unreported. min ls gains an ADDRESS column and --json carries the same facts. The EGRESS column prints the posture in the words min session policy uses, from one shared function, and the activate host line and the BOX NAMES row state how a box name opens through one resolver function. A VM node's host-address box reports no address until the node is granted one, since it is never mirrored onto 127.0.0.1. Backport of #2777. Adapted to the line: the policy.rs and create.rs hunks land in cmd/session.rs, and the minimald-rpc and minimald rpc tests land inline in lib.rs and rpc.rs, with the line's names (SessionStatus, EffectiveSessionPolicy, a built CreateSessionRequest). (cherry picked from commit 08045bf)
T110 —
min lsshows each box's address and egress posture, and states the naming surface as activate doesslice: S17
kind: task
covers: NET-142, NET-146, NET-147
size: L
est: 340 code, 500 tests
executor: agent
difficulty: routine
model: standard
diagnostics: what a person reads is the listing itself:
min lsandmin ls --jsonshow each box's address and egress posture, and the BOX NAMES row reads as the activate host line does; a box whose address or posture the daemon did not report shows-and its JSON omits the field, never an invented valueobservability: the daemon's existing ListSessions trace span records how many boxes it reported with an address and with a posture, so a listing that shows
-is visible in the tracelayers:
- listing-reply:
- crates/minimald-rpc/src/lib.rs: the ListSessions reply gains, per box and keyed by its id, the host loopback address its ports answer at (the NET-010 address, the
127.0.0.1fallback where the range is absent, the node's address for a host-address box) and its effective egress posture in the wordsmin session policyuses; both serde-defaulted so a daemon that predates them reads as unreported- crates/minimald-rpc/src/tests.rs: round-trips the new reply fields and reads an older reply without them
- crates/minimald/src/rpc.rs: fills the address and posture from each box's record, using the same effective-egress resolution
min session policyreads- crates/minimald/src/rpc/tests.rs: the listing reports a box's address and posture for own-address, host-address and none boxes
test: cargo nextest run -p minimald-rpc list_sessions_reply_carries_box_address_and_posture
test: cargo nextest run -p minimald list_sessions_reports_box_address_and_posture
- min-ls:
- crates/minimal/src/cmd/list.rs: the table gains an ADDRESS column after NAME showing the box's address (
127.0.64.2, or127.0.0.1where the range is absent,-when unreported);--jsoncarries it per session as"loopback_address": "127.0.64.2"; the EGRESS column prints the posture inmin session policy's words (deny-all (default),allow-all (default),deny-all,allow-all,declared rules), and for a host-address box whose rules this host cannot enforce per box it appends; per-box enforcement noneas the policy view does, neverper_boxor a bare-for a box whose egress is decided; the BOX NAMES row prints through the resolver's one surface statement below- crates/minimal/src/cmd/boxes/policy.rs: the posture words move into one function that both
min session policyandmin lscall, so the two surfaces cannot drift- crates/minimal/tests/cli.rs: builds the listing reply with the new fields wherever it spells the reply in full
- crates/minimal/src/tests.rs: pins the ADDRESS column, the JSON field and the EGRESS words against
min session policy's rendering of the same boxtest: cargo nextest run -p minimal ls_shows_box_loopback_address
test: cargo nextest run -p minimal ls_egress_matches_session_policy
- naming-surface:
- crates/minimal/src/resolver.rs: one function states how a box name opens on this host, and both
host_line_beside(activate) andname_surface_line(min ls) print it; where the box's ports answer at its own address the activate line readsnames: web.min.internal opens at its name on this machine; the HTTP proxy at 127.0.0.1:7654 opens it tooand the BOX NAMES row reads<box>.min.internal opens at its name on this machine; the HTTP proxy at 127.0.0.1:7654 opens it too; where they answer only at the127.0.0.1fallback both read… resolves on this machine; open it through the HTTP proxy at 127.0.0.1:7654; the proxy-only and proxy-not-serving arms keep one wording across both; the finish-setup clause is unchangedtest: cargo nextest run -p minimal ls_name_surface_matches_activate_host_line
test: cargo nextest run -p minimal activate_prints_one_host_line_per_install_state
test: cargo nextest run -p minimal activate_host_line_names_native_dns_when_deployed
test: cargo nextest run -p minimal activate_host_line_carries_finish_setup_clause_while_items_missing
produces: none
consumes: none
proof:
min lsshows each box's own address and the same egress posturemin session policyshows,min ls --jsoncarries the address, and the BOX NAMES row tells the reader how to open a box in the same words the activate line used.Proof
min lsnow has an ADDRESS column after NAME. It shows each box's own address (for example 127.0.64.2), 127.0.0.1 at the fallback, or-when the daemon did not report one.min ls --jsoncarries the address on each session as "loopback_address". The EGRESS column shows the posture thatmin session policyshows for the same box (deny-all (default),allow-all (default),deny-all,declared rules), with; per-box enforcement nonefor an unenforced host-address box. One resolver function writes both the BOX NAMES row and the activate host line, so they use the same words:<box>.min.internal opens at its name on this machine; the HTTP proxy at 127.0.0.1:<port> opens it too.Notes from the agent: 1) The spec excerpt does not contain NET-146 or NET-147. I used the task text as the contract for them. 2) scripts/session-e2e.sh lines 9415-9430 (outside the layers) still expect the old BOX NAMES words "resolves through this machine's DNS" and "the hostname proxy also listens on 127.0.0.1:". The task changes that row, so that KVM-lane e2e check will fail until a follow-up updates those two patterns to "opens at its name on this machine" and "; the HTTP proxy at 127.0.0.1:". The spec file docs/specs/18-spec-box-networking also quotes the old row text. 3) The activate host line cannot tell where a single box's ports answer. To pass that fact, create.rs (outside the layers) must change. So host_line_beside reads the own-address form for every native verdict. This is correct because a native verdict already needs the reserved range on this host's loopback. The 127.0.0.1-fallback form (
... resolves on this machine; open it through the HTTP proxy at ...) comes from box_name_opens(..., false). min ls uses it when every box address that the daemon reports is 127.0.0.1. 4) The task lists an unmarkedallow-allposture. No effective-egress resolution produces one (a declared section that allows everything prints as rows inmin session policy), so I did …[truncated]Tests
cargo nextest run -p minimald-rpc list_sessions_reply_carries_box_address_and_posturecargo nextest run -p minimald list_sessions_reports_box_address_and_posturecargo nextest run -p minimal ls_shows_box_loopback_addresscargo nextest run -p minimal ls_egress_matches_session_policycargo nextest run -p minimal ls_name_surface_matches_activate_host_linecargo nextest run -p minimal activate_prints_one_host_line_per_install_statecargo nextest run -p minimal activate_host_line_names_native_dns_when_deployedcargo nextest run -p minimal activate_host_line_carries_finish_setup_clause_while_items_missingVerify
min task run agent-verifyexited 100 under rustc 1.99.0, which the repository does not pin (1.99.1 in rust-toolchain.toml); the PR is not a draft on that alone — CI, which honours the pin, is the gate.Closes #2696