mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/delivery covered: a later merge that contains it was delivered: novox/mesh-host@4ba6d0864e3d (merged as d8ff1540 into main, walk plan-17915554870866…
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A service running as the operator — the tool runner, a unit of the
account's user manager — shares the operator's user id, and was as much the
terminal on the control-node as a person typing. The engine now reads the
asking process's cgroup and sends its login session's scope
(session-<id>.scope under the account's own user-<uid>.slice), or nothing;
the controller makes nothing else the terminal. The SDK is pinned at its
tag now that go/v0.1.9 exists.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
- Any failure of a request the link may have delivered now says the line
may have run; "nothing ran" is said only of a request the link knows it
never asked (no link open, nobody listening: link.ErrNotAsked).
- A test holds that a connection saying nothing is answered within its
deadline, set at accept, and gives its slot back.
- The README says the socket refuses a process outside the machine's own
namespaces, so flatpak, toolbox and distrobox shells are refused.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
- A request that timed out was said as "the controller was not reached,
nothing ran", though the line may have run. The engine now follows a line
the controller answers as running, by its call, until it ends; when no
answer comes in time it says the line may have run and names the call.
"Nothing ran" is said only of a request no controller took.
- One account could hold the socket: accepting waited for a free slot, and
a connection could stall before its deadline was set. The deadline is set
at accept, before any lookup; each user id has two lines at once and the
machine eight, and one over is answered busy at once.
- A container running with the operator's user id shares the number, not
the machine: a peer whose mount or user namespace is not the engine's is
refused, and nothing is asked.
SDK pinned at the go/cli branch's newest commit until go/v0.1.9 is tagged.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The homes come from a map of the user database, and the test compared them in
the order it happened to iterate: it failed one run in four, the repo check
of 2026-10-09 among them.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The search kept a two-minute limit of its own inside its 15-minute bound, so
a long walk always ended incomplete; pacman was asked once per path; and find
exiting with an error but some output was read as a complete walk. The bound
and the not-judged reasons move to the public rootsearch package so the
controller reads the same values (the review of 2026-10-09).
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The records send the operator to "the controller's terminal", and nothing
reaches it (hq issue 343). mesh-cli holds no credential: it asks the engine
on its own machine over a socket. The engine reads the asking account from
the kernel (SO_PEERCRED), never from the program, and asks the controller on
mesh.control.<node>.cli, its own node's subject, so the bus server says which
node asked. The controller decides what the line may do; nothing received
on the socket is run or applied here, and the socket has no network address
(ADR 0004 stands).
The local protocol is the SDK's go/cli package, at a pseudo-version of its
branch until that merges and is tagged go/v0.1.8.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A search over a large disk ran inside the look, holding the judge's lock and
stalling the health statement; it now runs apart, serving its last result,
backing off after a failure, and saying not judged until it has one. And a
rule naming a user, or a comment, no longer hides another rule's
unconditional yes (hq ADR 0266).
Issue 339's guard (#63/#64/#65) now refuses the machine's own trees, another account's
directory below a home, and uses a found directory as found; this branch's own copy of
those rules goes. What stays is the one rule only an account that never becomes root
needs: below its home it owns every parent and could swap a link in between the
engine's check and an unpack or a write-into (hq ADR 0266 §7). One reader of the user
database serves both.
A hard link swapped in for ~/.claude would have had root chown another
account's file; fstat on the descriptor now refuses a second link, a fifo or
an unexpected kind before anything is changed (hq ADR 0266, the re-review).
The judge also finds polkit rules for every account, a runtime's API on TCP,
setgid-to-root programs whoever owns them, setuid programs on every suid
filesystem, and unprotected links; the rest is listed as not judged.
The controller cannot read a machine's user database, so a name it is given
for the agents' account (hq ADR 0266) may be a service's own; taking it over
would hand agents that service's files. Refuse it, touching nothing.
An account could replace ~/.claude with a link to /etc and have the
node-engine chown, chmod or write through it on the next apply. Below a
person's or an agent's home every component is now opened without
following a link, and a link refuses the resource in words.
The root judge also reads doas and polkit rules, the container runtimes'
sockets with their ACLs, ACLs on the secrets, and setuid-root programs no
package owns, in the C locale; Judged and NotJudged write down exactly
what it covers (hq ADR 0266 review).
An account an agent runs as (novox/hq ADR 0266) is read on every look for a
uid of 0, a group that grants root, any sudo rule and a mesh secret it can
read; any way found is unhealthy and said, a read that fails is unknown.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
#64/#65 and #66 each added Applied.Declared: a directory's declared owner and
mode (string) and a service's held state (*FoundUnit), both under the JSON key
"declared". The directory's is already on nodes, written by d28ca95, so it
keeps the name and key; the service's becomes DeclaredUnit, "declared_unit".
#66 never deployed, so no node holds a record under the new key.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
Third review of mesh-catalog #147.
A unit's name can now come from a setting (a template's instance), so every systemctl call that names a
unit passes `--` before it, and a name can never be read as an option.
On removal, a unit is started or enabled again only when it is exactly the stopped and disabled the mesh
held it to. A masked unit, or one somebody started or enabled since, is left as it is, and the report
says why. A restore that fails (the start refused, or the unit not running after it) puts back what this
attempt changed and is reported as waiting, so no apply is made fatal by it and the next one tries again.
The tests' fake service managers read systemctl's arguments in the shape they were written for
(oldShape); internal/system's own test holds the `--`.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The fourth review (hq issue 339): `trusted` is the catalogue's word and the
controller takes it out, but a controller that passes it on, an older one or
one rolled back to, must not cost a node its whole declaration. A file now
accepts it and nothing reads it; any other resource still refuses it. And a
podman that hangs no longer eats docker's ten seconds.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A unit found running and enabled that a declaration held stopped and disabled stayed stopped and
disabled when the declaration went: ADR 0118 never starts a unit on the way out. Unassigning the zfs
module would so have left the home server's pool with no scrub schedule at all, said nowhere.
The host now records what it declared for a service (Declared, beside Found). On removal, a unit found
running and declared stopped is started again, and one found enabled and declared disabled is enabled
again, each read back; the report says restored, and the plan says it first. A record from before the
field starts nothing.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The third review of #64 (hq issue 339): an ExecStart continued over lines, a
quoted path, a variable from Environment= or EnvironmentFile= across the unit
and its drop-ins, dockerd's --config-file and podman's TOML literal strings
each hid a data root from the guard. The running runtimes are now asked first
(docker info, podman info) on a runner of the guard's own, and the files are
read beside them in all those forms.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/delivery covered: a later merge that contains it was delivered: novox/mesh-host@7c046a8a1d84 (merged as d28ca959 into main, walk plan-17915049347305…
The review of #64 asked for containerd's tree, dockerd's --data-root from its
unit and drop-ins, and podman's graphroot to be refused like docker's own
(hq issue 339); and the controller now tells a directory a send found from
one found before it, which needs the time it was first found kept across
applies instead of the time of the last apply. A hand-over now names the
resource it was for, so it lapses if another declaration points at the path.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A directory the engine used as found kept its owner and mode in silence, so
a module that cannot write it would just fail (hq issue 339 review). Each
one is now a WARNING in every apply's log and its module's resource of kind
directory, stated unhealthy on every look with its owner and what was
declared, which the controller raises as the module's condition; `mesh-host
hand-over <dir>` at the machine hands it to the mesh, and the next apply
gives it the declared owner and mode. The guard also refuses the spool, /opt,
the container runtimes' data (its configured data-root too), and an account's
.ssh as an access, a mount, or a directory anywhere but its own home.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A module's places setting could point a directory anywhere, with an owner it
named, and the engine chowned and chmodded whatever it was pointed at as root,
through a link too; its accesses setting could mount any path into a container
(hq issue 339). The engine now refuses, as a failed resource with its reason,
a directory, access or mount source that is or holds one of the machine's
roots, anything placed in the kernel's, the boot loader's, root's or the
engine's own trees, a non-root directory below /etc, /usr or /run, an access
there, another account's directory below a home, and a container mounting
anything it refused, judged with every link resolved. A directory whose last
component is a link is refused, and one found here that the mesh did not make
keeps its owner and mode and is never removed.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/repo-check pass: THE CHANGE ALTERS ITS OWN CHECK (merge-check.sh): main's version judged it; the change's judges the pull requests after it merges; it…
The evidence read "asking it: asking <subject>": the words are now the link's own.
A deadline is said as the time the check gave it. A refusal is said only when the bus
refused this question, not an earlier one under a grant since widened. merge-check.sh
runs the tests that need a bus against a throwaway nats-server when the toolchain has
one, and says so when it does not (hq issue 331).
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
The queue asks a tool through a ToolBus assertion that only OverNATS satisfied; the
link the engine actually holds (natsLink) had no Ask, so every tool check failed as
"no link to the bus is open" while the link was up (hq issue 331). Give natsLink Ask,
assert at build time every optional interface the queue expects of it, and tell a link
that cannot ask apart from no link.
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
An APU's integrated graphics and Intel's second (MMIO) view of a package made a machine
report a power meter the sampler would not read, so its bar could only say not
measured.
The power module is on every machine and its bar blocks apply only where the
hardware is. Two capabilities read from the kernel let the controller compose them
there and nowhere else: battery (a Battery supply not scoped to a device) and
power-meter (a system battery, a powercap package counter, or a graphics device
reporting its power).
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
mesh/merge-gate pass: builds mesh-host → ace, g14, novox, shanks; no bus step; every machine composes with the change as it did without (4 of 4 compose)
A module puts the operator's account in a group by declaring the account with that group alone.
The node-engine now records each group it added, takes back only those when nothing declared still
asks for them, refuses a group the machine lacks before usermod runs, and states each such account
as its module's resource of kind account: relogin needed while the running session lacks the group.
jochen
mesh-admin