Self-review catch. The ladder reported "no tea login named X exists" when tea
was simply absent -- a cause it never checked, sending the reader off to create a
login they have no tool to create. Same defect class as the one just filed as
#1357, so shipping a fresh instance of it in this PR would be poor form.
get_gitea_login_for_host() returned the FIRST tea login matching the host. With
43 logins on this host, roughly half match one server, so a seat whose own login
was missing silently acted as whichever identity happened to sort first. That
satisfies gate 16 mechanically (an author and a reviewer exist) while violating
it (both are the same actor under two names).
A seat now declares itself via MOSAIC_GIT_IDENTITY or `git config
mosaic.gitIdentity`, and resolution derives the canonical login name from that
identity plus the instance (`<instance>-<seat>`). If that login is absent it
fails closed with a named error and the command to create it. It never borrows.
Same rule on the --repo override path, which had it worse: it fell through to
get_default_tea_login(), i.e. the default-marked login or, failing that, the
first login of ANY host -- an identity chosen by config file order. The four
callers now pass the owner so the instance can be derived. With no identity set
(a human at a terminal) the old fallback is unchanged, which is the same point
at which the token path stops enforcing.
lane-brief.sh mapped owners straight to the SHARED `usc` / `mosaicstack` logins.
The ladder now goes first there, and a seat that cannot resolve its own login
exits rather than falling through to a shared one.
Also adds tools/fleet/seat-logins.sh: projects seat credentials into tea logins
under canonical names, so the name this code requires is one an operator can
mechanically produce rather than hand-maintain.
Test notes:
- The suite had TWO sandbox helpers, run_in_repo and a near-copy run_in_repo2.
The copy drifted: it never got the identity unset, so the suite kept failing on
a provisioned seat after the original was already fixed. run_in_repo2 now
delegates, so the guarantee lives in one place.
- New coverage for both ladder branches (login present -> used; absent -> named
error and NOTHING on stdout, proving it did not borrow the matching login
sitting right there), both identity rungs, the --repo path, and explicit
GITEA_LOGIN outranking the ladder. Each verified by injecting the regression it
claims to catch and confirming it goes red.
- test-issue-create-body-safety.sh now pins the no-identity case; its subject is
body quoting, and an ambient seat identity made it fail for an unrelated reason.
- test-issue-close-fail-closed.sh derives its fixture login from the runner's
identity. This does not make it hermetic and does not claim to: its API-path
cases need a real credential for the runner's own identity, so it passes only
where the runner owns one, on this branch and on its base alike. Pre-existing,
documented in the PR rather than papered over.
Two changes to one rule: a credential is resolved from exactly one place,
and an identity that cannot be resolved is refused rather than substituted.
FAIL CLOSED. Both readers ended in an unconditional fall-through to the
shared Gitea account whenever an identity did not resolve. Every seat in a
fleet therefore pushed, opened PRs and filed reviews under one account, and
a record made that way cannot be traced to the agent that made it
afterwards. The fallback now applies only where there is no attribution to
lose: a host with no fleet. Where seats exist, an unresolvable request emits
nothing, exits nonzero, explains itself on stderr, and — in the git helper —
appends a record naming the identity, host, reason and cwd, and no token
value, to ${MOSAIC_CREDENTIAL_SPOOL:-~/.local/state/mosaic-credential-escalations}.
A host runs a fleet when <brain>/fleet/agents exists, which is the signal
packages/mosaic/src/fleet/brain-home.ts already uses to decide a brain is
active, resolved the same way (MOSAIC_BRAIN_HOME, else ~/.mosaic). This is
what keeps the change a no-op for an operator who has not provisioned
per-slot tokens: no fleet directory, shared account, unchanged. It is also
why there is no environment variable to restore the old behavior — one would
reintroduce the substitution being removed.
STORE SELECTION. Both readers hardcoded ~/.config/mosaic/secrets/gitea-tokens,
so a seat's own secrets/ slot was invisible to the framework: a seat could
hold a valid credential and still be served the shared account. The store is
now chosen by what the identity is. An identity with a directory under
<brain>/fleet/agents/ is a seat and is read only from
<brain>/fleet/agents/<id>/secrets/; any other identity is a service identity
and is read from the framework store. There is no precedence between them
and no fallback from one to the other, so a seat with an empty slot is
refused even when a same-named token sits in the framework store. Two copies
of one credential are drift rather than redundancy, and drift surfaces as
the stale copy returning 401, which reads as a revoked token and sends
whoever debugs it somewhere else.
detect-platform.sh is in scope alongside git-credential-mosaic because they
are the two readers of these tokens. Patching only the git helper would make
"one credential, one location" true for push and fetch and false for
pr-create.sh, issue-create.sh and pr-review.sh, which is the harder failure
to notice.
TESTS. The three assertions that pinned the shared-account fall-through are
now fail-closed assertions, and a refusal is checked four independent ways:
nonzero exit, empty stdout, a stderr diagnostic naming identity and host,
and no shared token value anywhere in the output. The exit code alone would
pass against a helper that emitted the credential and then failed. Added:
seat-slot resolution, the no-cross-store-fallback case with a control
proving the framework-store file it declines to read is readable, no-identity
on a fleet host, the fleet gate firing on the default ~/.mosaic and not only
on an injected MOSAIC_BRAIN_HOME, and a cross-host leak check. Both suites
were run against the pre-change code as a control and fail there on exactly
the shared-token emission.
shellcheck is not installed on the authoring host, so the rewritten helper
is unlinted locally and CI is the first lint of it.
fred
ops-01
jason.woltjeandMos