Advisory codes
Every advisory retrofit prints, by code. The register is the word a text line opens with and
the JSON level: warning or note. A braced name such as {table} is a value the printed
line fills in. A message that varies with a field shows each form on its own line, led by the
field’s value. Where the message is only such a value, its detail and hint lines follow it.
| Code | Class | Register | Message |
|---|---|---|---|
adopted-default-data-root | schema | warning | adopting the default data corpus root {path}, which already holds data files that no schema of this project owns, or that cannot be read as this project’s |
after-traffic-outcome | data-motion | note | {operator} {opid} meets rows that traffic wrote after it |
already-pinned | dml | note | {path} already anchors on {opid}; nothing was written |
apply-risk-add-exclude | apply-risk | warning | adding an EXCLUDE constraint builds its index and checks every row under an exclusive lock, and PostgreSQL has no concurrent form; on a populated table run it in a window that can hold the lock |
apply-risk-add-not-null | apply-risk | warning | SET NOT NULL scans the table under an exclusive lock; on a populated table add a NOT VALID CHECK (col IS NOT NULL) in an earlier migration, then VALIDATE it and SET NOT NULL to avoid the long lock |
apply-risk-add-unique | apply-risk | warning | adding a UNIQUE constraint builds its index under an exclusive lock; on a populated table CREATE UNIQUE INDEX CONCURRENTLY first, then ADD CONSTRAINT … USING INDEX to avoid the long lock |
apply-risk-data-corpus | apply-risk | warning | {table}: some rows are not owned by this schema |
apply-risk-data-dropped | apply-risk | warning | history-bound, unroutable-purpose, not-selected, superseded, table-dropped, table-filtered, undeclared, table-joined, table-split, table-merged, table-inlined, table-unjoined, reference-data: {table}: {exclusion_cause}: some rows cannot be applied by this runddl-only: {table}: some rows are excluded from the captured data |
apply-risk-data-exclusions | apply-risk | warning | history-bound, unroutable-purpose, superseded, table-dropped, table-filtered, undeclared, table-joined, table-split, table-merged, table-inlined, table-unjoined, reference-data, ddl-only: {table}: some rows are excluded from the captured datanot-selected: {count} dataset(s) not named ({table_count} table(s)); name them with --label, or run retrofit data list to see what is present |
apply-risk-data-exclusions-table | apply-risk | note | history-bound, unroutable-purpose, not-selected, superseded, table-dropped, table-filtered, undeclared, table-joined, table-split, table-merged, table-inlined, table-unjoined, reference-data: {table}: {exclusion_cause}: some rows did not reach the emitted bandddl-only: {table}: some rows are excluded from the captured data |
apply-risk-data-not-reversed | apply-risk | warning | down-not-reversed: {table}: {down_cause}: preserved DML is not reversed by --downno-fold-at-target: {table}: {down_cause}: the reference table has no fold at the target; its rows are kept |
apply-risk-database-at | apply-risk | note | schema {schema}: apply-risk is judged from the database cut {opid} (pinned by {name}) |
apply-risk-database-at-behind-seal | apply-risk | note | schema {schema}: apply-risk is judged from the database cut {opid} (pinned by {name}), which precedes the latest Seal; tables created, and names given, after it and at or before that Seal are judged new, which holds only if the database’s next run goes from the pin past this op |
apply-risk-database-at-ignored | apply-risk | note | the database-at pin is ignored here, which resolves none: {name} |
apply-risk-exclusive-lock | apply-risk | warning | {operator} {opid} moves rows under an EXCLUSIVE lock on {table} held to COMMIT, blocking writes and foreign-key checks for the whole move |
apply-risk-import-summary | apply-risk | warning | the import carries {count} apply-risk advisories |
apply-risk-matview-refresh-cost | apply-risk | warning | {table}: refreshing this materialized view is not free |
apply-risk-partition-tree-locks | apply-risk | warning | {operator} {opid} locks over 1,000 relations in one transaction |
apply-risk-recorded-unsealed | apply-risk | note | names recorded into the unsealed tail are read by apply-risk as unreleased and new |
apply-risk-rename-in-use | apply-risk | warning | table: renaming table “{name}” breaks readers still using the old namecolumn: renaming column “{name}” breaks readers still using the old name |
apply-risk-share-row-exclusive-lock | apply-risk | warning | validating foreign key on a populated table acquires SHARE ROW EXCLUSIVE lock for the scan, blocking writes |
apply-risk-table-rewrite | apply-risk | warning | CastColumn {opid} changes the type of {columns} on {table} under an ACCESS EXCLUSIVE lock |
apply-risk-validating-check | apply-risk | warning | validating CHECK constraint scans and locks the table; author it as NOT VALID (not_valid=true) then VALIDATE CONSTRAINT in a later migration (separate transaction) to avoid the long lock on a populated table |
apply-risk-validating-fk | apply-risk | warning | validating FOREIGN KEY scans and locks the table; author it as NOT VALID (not_valid=true) then VALIDATE CONSTRAINT in a later migration (separate transaction) to avoid the long lock on a populated table |
apply-risk-view-rebuild | apply-risk | warning | ReplaceView of {target} renders as a rebuild: {reason} |
apply-risk-window-add-not-null | apply-risk | warning | schema {schema}: op {opid} sets NOT NULL on a column of {table} inside the migration window; the scan runs under an exclusive lock on a table that may hold rows |
apply-risk-window-unsealed | apply-risk | note | schema {schema}: no Seal, so the migration window begins at root and {table_count} tables in it were treated as newborn (empty); seal what production carries so the apply-risk gate can see it |
apply-risk-window-validating-check | apply-risk | warning | schema {schema}: op {opid} validates a CHECK constraint on {table} inside the migration window; the scan locks a table that may hold rows |
apply-risk-window-validating-fk | apply-risk | warning | schema {schema}: op {opid} validates a FOREIGN KEY on {table} inside the migration window; the scan locks a table that may hold rows |
apply-risk-withheld-newborn | apply-risk | note | newborn: schema {schema}: apply-risk withheld for table {name}, created at {opid} and not populated since (judged from {reason})rename: schema {schema}: apply-risk rename withheld; old name {name} is absent at F (judged from {reason}) |
atomicity-segment-boundary-crossed | atomicity | warning | {opid}: this migration spans more than one transaction |
atomicity-wrapper-suppressed | atomicity | warning | no BEGIN;/COMMIT; wrapper was emitted for this file |
boundary-requirement-commit-boundary | atomicity | warning | {operator} needs a transaction boundary that --envelope={target} leaves to the apply context |
boundary-requirement-non-transactional | atomicity | warning | {operator} needs a transaction boundary that --envelope={target} leaves to the apply context |
cascade-non-atomic-on-no-transaction | atomicity | note | ChangePrimaryKey appended with N={count} dependent FK rewrite(s); cascade must run inside one transaction to apply atomically (the default --envelope=own wrap satisfies this; --envelope=caller and --envelope=independent emit no wrapper, so the apply context must supply one) |
change-primary-key-non-atomic | atomicity | warning | cascade may apply partially: no BEGIN/COMMIT wrapper was emitted under --envelope={target} |
column-narrowing-breaks-dependent | schema | warning | {target} reads {columns} of {name}, which the ReplaceView no longer holds, so the migration’s re-create of {target} fails at apply |
corpus-file-skipped | schema | warning | skipped malformed corpus file {path} |
corpus-foreign-records-present | apply-risk | warning | unregistered-owner: {count} record(s) under this corpus root are owned by other schemas: {owning_schemas}zero-owned: no record under this corpus root is owned by this schema; all {count} belong to: {owning_schemas} |
corpus-records-owned-by-other-schema | apply-risk | note | other-schema: {table}: {count} record(s) under this corpus root are owned by schema {schema}unregistered-schema: {table}: {count} record(s) under this corpus root are owned by no registered schema |
corpus-rollback-incomplete | schema | warning | rollback of {name} failed |
data-file-missing | data-motion | warning | the data record for {table} at {opid} ({path}), in the range this down reverts, is named by a witness, but no file carries it |
data-file-unaddressed | schema | warning | {path} is not an addressed data file and was not checked |
data-file-unfilled | schema | warning | unfilled data file {path}: it declares body=data but holds no records, or a record still carrying scaffold placeholders |
data-layer-below-floor | dml | note | {label} contributes to {table} at {contribution-anchor}, below the {lower-floor-kind} body={lower-floor-body} floor {lower-label} placed at {lower-floor-anchor} |
data-loss-on-apply | schema | warning | {reason} hint: matview-cache, null-key-rows, unreferenced-lookup-rows: capture the affected rows before applying, or re-shape the change to preserve themhint: drop-or-cast: capture the affected rows before applying, or re-shape the change to preserve them; under migrate --down, restore them from a backup taken before that op applied |
data-loss-on-apply-narrowing-cast | schema | warning | {table}: narrowing cast from {from_type} to {to_type} may lose data on apply |
data-loss-on-apply-render-chosen | schema | warning | the reverse of {operator} {opid} loses the rows of {name} ({grade}) |
data-motion-cascade-scope | data-motion | note | {reason} |
data-motion-copy-post-fork-independence | data-motion | note | the copy and {table} are independent identities from this op forward; post-op edits to either table do not propagate to the other |
data-motion-declaration-inherited | data-motion | note | {table} is reference data because its source {target} is declared; if its rows are not reference data, withdraw it with retrofit schema ref-tables remove --schema {schema} {table} |
data-motion-deferred-split | data-motion | note | {table}: {count} split candidates deferred to a later batch |
data-motion-fresh-sequences | data-motion | note | {reason} |
data-motion-keyed-fold | data-motion | note | {operator} {opid} folds each inlined source’s columns and rows into {table} along their primary keys and drops the sources; the inlined columns take the survivor’s privileges and an inlined source’s grants end with it |
data-motion-partition-not-pg-declarative | data-motion | note | SplitTable is a row-axis split that moves rows into new tables and keeps its source; it does NOT emit PARTITION BY / PARTITION OF declarative-partitioning DDL |
data-motion-product-capture-owed | data-motion | note | {table} is reference data (declared source {target}), and the rows this op gives it exist only in migrated databases: capture {table} at HEAD (retrofit data capture <dump> --table {table}, then retrofit data witness) before op seal, which refuses reference-rows-uncaptured until then |
data-motion-replica-column | data-motion | warning | a snapshot proof writes {columns} into every row of {table}, so a logical-replication subscriber whose copy lacks the column stops applying at this transaction |
data-motion-source-consumed | data-motion | note | {count} source(s) end at this op; {table} stays and takes their rows; snapshot DDL stops emitting the sources from this opid forward |
data-motion-triggers | data-motion | note | {operator} {opid} writes rows the triggers listed here fire on; the render toggles no trigger |
data-only-single-statement-autocommit | atomicity | note | this --data-only band applies as individually autocommitted statements at {opid}: an abort mid-band leaves every completed phase committed |
data-qualified | dml | note | record seq={seq}: the renderer qualified {qualified_sites} |
data-replace-cascade | dml | warning | reaching: replacing {table} at {opid} reaches into {name}, which this slot does not replace ({reason})blocking: replacing {table} at {opid} may be blocked by rows in {name}, which this slot does not replace ({reason})reaching_ref: replacing reference table {table} at {opid} reaches into {name}, which this slot does not replace ({reason})blocking_ref: replacing reference table {table} at {opid} may be blocked by rows in {name}, which this slot does not replace ({reason})cross_slot: replacing {table} at {opid} acts on {name}, which is replaced in a different slot of this script ({reason})keyed_child: clearing {table} at {opid} meets rows in {name}, whose keyed staged block retains them through the clear ({reason})self: replacing {table} at {opid} acts on its own self-referencing foreign key ({reason}) |
data-replace-crossing | dml | warning | reference table {table} (floor at {opid}) is written as a delta because this range crosses the removal of its declaration at {name}; rows the corpus never held are kept ({reason}) |
data-replace-prologue | dml | warning | {table} is replaced by its full observed state at {opid} ({reason}) |
data-replace-unordered | dml | note | the foreign-key graph could not be read at {opid} ({reason}), so the prologued tables ({values}) are emitted in the stated tie-break order |
data-replace-whole-reload | dml | warning | reloading {table} whole; {values} are not carried by the body and will be set to NULL |
data-statement-order-fk | dml | warning | state-before-statement: the rows that state {table} load before the statements that supply {target} ({name})child-statements-first: {table}’s statements run before {target}’s under {name} |
data-witness-mismatch | data-motion | warning | the data record for {table} at {opid} ({path}), in the range this down reverts, no longer matches its witness |
data-witness-missing | data-motion | warning | the data record for {table} at {opid} ({path}), in the range this down reverts, has no witness in the covering seal’s range |
database-at-op-revised | apply-risk | note | schema {schema}: op {opid} at or before the database-at pin is no longer held as applied; the pinned database applied it (a working copy to rebuild) |
database-at-unresolved | apply-risk | note | schema {schema}: the database-at pin {reason} (from {name}) does not resolve; the declared database’s cut is unread, so apply-risk withholds nothing |
deferred-keys-bracketed | data-motion | note | the script sets the INITIALLY DEFERRED foreign keys of the ops that write rows immediate for each op and returns them to deferred |
deferred-keys-owed | data-motion | warning | the script writes rows under INITIALLY DEFERRED foreign keys and sets no constraint mode; under --envelope caller the enclosing transaction owes it |
deferred-keys-pending | data-motion | warning | data record seq={seq} leaves foreign-key checks on {schema}.{table} queued when op {operator} {opid} alters the table; that op fails 55006 (pending trigger events) |
deferred-trigger-owed | data-motion | warning | the script’s writes leave events of the deferred constraint trigger {schema}.{name} queued on {table} when op {operator} {opid} alters the table; that op fails 55006 (pending trigger events) |
deferred-trigger-pending | data-motion | warning | data record seq={seq} leaves events of the deferred constraint trigger {schema}.{name} queued on {table} when op {operator} {opid} alters the table; that op fails 55006 (pending trigger events) |
dml-behind-window | dml | warning | root-boundary: DML seq={opid} purpose={purpose} anchored on the pre-genesis root boundary is behind --from; omittednamed-opid: DML seq={opid} purpose={purpose} anchored on {name} is behind --from; omittedsealed-opid: DML seq={opid} purpose={purpose} anchored on {name}, inside seal {target}, is behind --from; omitted |
dml-body-meta-command-admitted | dml | warning | {name} was emitted into the artifact under --allow-unscreened-dml; psql runs it client-side, outside retrofit’s transaction framing — a session command such as \connect ends the transaction open before it |
dml-body-partition-dependent | dml | warning | {name} cannot run inside a transaction block when its target relation is partitioned, and retrofit does not read the database catalog; if the target is partitioned, this window applies as one more segment than reported |
dml-chain-near-miss | dml | warning | {reason} hint: re-run with the prior label-set (or hand-edit the labels) if this was meant to supersede |
dml-dump-band-listed | dml | warning | {count} rows listed, not captured: {table} |
dml-op-show-matview-replace | dml | note | replacing materialized view {name} discards its stored rows |
dml-purpose-unclassified | dml | note | transition: {opid}: DML is classified as a transitionseed: {opid}: DML is classified as a seedcleanse: {opid}: DML is classified as a cleanseunclassified: {opid}: DML carries no purpose classificationref: {opid}: DML is classified as reference data |
dml-ref-table-flag-ignored | dml | note | {name} ignored: {table} is a declared reference table, so its record is purpose=ref |
dml-ref-table-no-match-universe | dml | note | schema {name} has no ops at HEAD, so no pattern can match yet |
dml-ref-table-policy-disagreement | dml | warning | {count} tables carry committed records their reference-data declaration would classify differently |
dml-ref-table-seed-dormant | dml | note | seed[{seed-labels}] holds {table} ({opid}); kept, never reclassified, and dormant while {table} is reference data |
dml-seed-band-opt-in-required | dml | warning | records in a purpose=seed file apply only{reason}; without it they are excluded and counted |
dml-superseded-in-window | dml | warning | {name}: a superseded reference record is needed by a step before the floor this run mints; the window from the last Seal will refuse |
dml-unclassified-listed | dml | warning | {count} records unclassified: {table} |
drift-convert-recognition-gated | schema | note | partition-convert recognition inactive |
drift-extract-unrecognized | schema | warning | extract-shaped reshape not recognized as an ExtractTable |
drift-fk-facet-recapture | schema | warning | this recreate reconciles referential-action facets that differ between the oplog and the database; capturing them converges in one pass and does not recur |
drift-merge-shape-mismatch | schema | warning | merge sources differ on a compared field |
drift-move-column-candidate | schema | warning | a column dropped from one table and added to another was not recognized as a MoveColumn |
drift-skip-internal-defect | schema | warning | drift skip {skip_kind} on {name} in schema {schema} (line {line}) is a retrofit defect: valid PostgreSQL does not reach it with the pinned parser |
drift-skip-unclassified | schema | warning | unclassified drift skip {skip_kind} on {name} in schema {schema} (line {line}): {reason}; retrofit has no class for this skip, so it counts as a gap |
drift-split-unrecognized | schema | warning | same-shape table beside a surviving source was not recognized as a SplitTable |
drift-survivor-extract-unrecognized | schema | warning | column extract off surviving table {table} was not recognized as an ExtractTable |
drop-opaque-readers | schema | warning | table: dropping table “{name}” leaves function bodies retrofit cannot read; any may still use itcolumn: dropping column “{name}” leaves function bodies retrofit cannot read; any may still use itfunction: dropping function “{name}” leaves function bodies retrofit cannot read; any may still call it |
element-ref-absent | schema | warning | the element {element_id} has no ref that binds it under every operator that takes its kind |
enum-value-commit-boundary | type | note | the new value {name} on {target} is not usable until the adding transaction commits |
excluded-schema-referenced | schema | note | kept statements reference objects of the excluded schema {name}; each keeps the disposition a reference into an unmanaged schema has |
exclusion-applied | schema | note | ~ (excluded) {excluded_kind} {name} — {count} statement(s) — {reason} |
experimental-graduated | schema | warning | [experimental] purposes lists {name}, which is no longer experimental |
experimental-purpose | dml | warning | {path}: purpose={purpose} records are experimental: a released record is protected only by its position in the log, not by a Seal’s digest |
extract-rejoin-key-added | data-motion | note | the extract added rejoin keys that appear in no oplog text |
extract-rejoin-key-no-delete-action | data-motion | note | {operator} {opid} ties each child to {table} by a rejoin foreign key with no ON DELETE action, and its forward gives every {table} row a child row: a {table} row that has one can no longer be deleted |
extract-reverse-guard-non-atomic | atomicity | warning | a guard in this script is not atomic with the changes before it: no BEGIN;/COMMIT; wrapper was emitted under --envelope={envelope} |
extract-reverse-ordinal-divergence | data-motion | note | extract-reverse-ordinal-divergence |
extract-survivor-arm-selected | data-motion | note | source table {table} survives this extract in place |
from-dump-seal-run | dml | note | the dump matches a run of seals joined only by ops not observable in a schema-only dump |
gitattributes-missing | schema | warning | .gitattributes is missing the {config_path} line |
import-force-orphaned-references | schema | warning | schema {schema}: {count} bodies reference elements no registered schema resolves after this --force: {values} |
internal-advisory-write-failed | internal | warning | an advisory could not be written to its channel: {reason} |
internal-invalid-advisory | internal | warning | zero-def: advisory definition {original_code} is incomplete and cannot describe itselfundeclared-field: advisory definition {original_code} was supplied a field it does not declaremissing-field: advisory definition {original_code} is missing a value for a field it declarescontainer-mismatch: advisory definition {original_code} was supplied a value at the wrong container kindoff-token-discriminator: advisory definition {original_code} was supplied a discriminator value outside the closed token set it declares |
internal-reserved-type-violation | internal | warning | record type {type_name} is reserved for the advisory carrier and cannot be emitted through the generic {entry_point} writer |
join-key-demoted | schema | note | JoinTable carries primary key {name} ({element_id}) onto {table} as UNIQUE: {reason}; kept key {target} |
label-duplicated | dml | note | label {label} was named more than once; the repeat was dropped and the layering is unchanged |
label-unmatched | dml | warning | label {label} matched no data records; nothing was rendered for it |
legacy-origin-default | schema | note | a column’s stored default is a different spelling of the authored default |
migrate-segmented-non-atomic | atomicity | warning | {reason} hint: apply it with a runner that does NOT open its own transaction (plain psql -f); under --envelope=own a runner wrapping the file in one transaction is refused by name at the head (own-envelope-transaction-open) before any DDL |
move-column-down-refused | data-motion | note | migrate --down refuses a window holding this MoveColumn (migrate-move-column-needs-default): the moved column is NOT NULL with no DEFAULT |
owned-sequence-name-stale | schema | warning | a skipped statement leaves the owned sequence name {name} of {table}.{column} stale |
privileges-not-restored | schema | warning | view-replace: view {name} is dropped by the preceding DropView and created again by this CreateView: privileges on {name} are not restoredunjoin-child: the UnjoinTable child {name} takes its source’s name: the source’s privileges on {name} are not restored onto it |
reference-chain-unfoldable | dml | warning | reference chain {table} does not fold offline: {reason} |
reference-declaration-unjudged | data-motion | note | op {opid} combined declared and undeclared tables into {schema}.{table} ({values} not declared), and no judgment records whether it is reference data; record one with retrofit schema ref-tables add --schema {schema} {table} or retrofit schema ref-tables remove --schema {schema} {table} before op seal, which refuses reference-declaration-unjudged until then |
reference-key-lost | dml | warning | reference table {table} lost {count} key(s) since its chain: {name} |
reference-restate-authored | dml | note | replace: reference table {table}: a hand-written restate states the whole table and renders as a replacebodies: reference table {table}: a hand-written restate holding a statement other than an INSERT keeps its bodies: migrate renders it as written, not as a replace, and applying it again can fail on rows it already wrotemint: reference table {table}: a hand-written restate states the whole table and, while its block holds INSERTs only, renders as a replacelayered: reference table {table}: a hand-written restate under a label renders as a replace only while its block holds INSERTs only and no lower layer migrate selects holds the table; otherwise it renders as written, and applying it again can fail on rows it already wrote |
reference-table-emptied | dml | warning | reference table {table} was observed empty: the replace body=none removes every row |
rename-opaque-readers | schema | warning | table: renaming table “{name}” leaves function bodies retrofit cannot read; any may still use the old namecolumn: renaming column “{name}” leaves function bodies retrofit cannot read; any may still use the old name |
render-skip-element-not-rendered | render-skip | warning | view-body: {name} {element_id} omitted from snapshot: its body cannot be reconstructed without a live catalogelement-body: {name} {element_id} omitted from snapshot: its body references an element not live as of the cutunsupported-node: {name} {element_id} omitted from snapshot: an unsupported node in its body could not be rendered |
reverse-declared-noop | atomicity | note | RunRawScript {opid} contributes nothing to the down: it declares no down action is needed ({reason}) |
reverse-guard | atomicity | note | check: {operator} {opid} returns the check on {name} as a guardrule: {operator} {opid} returns the rule on {name} as a guardreference-edge: {operator} {opid} returns the reference edge on {name} as a guard |
reverse-irreversible | atomicity | note | paired: {name} is paired, not the element the target held: its rows left with the dropnot-null-not-restored: {name} comes back nullable: its NOT NULL is not restoredholder-child: {name} is rejoined as the holder child, not the element the target heldmatview-unpopulated: {name} is re-created WITH NO DATA, unpopulated until its REFRESHsynthesized-name: {name} is synthesized from a new ElementID, not the name the target’s cut rendersre-minted-name: {name} is re-minted by the key change, not the name the target’s cut rendersorder-not-restored: {name} keeps the column order the down leaves: its order is not restoredidentity-absent-not-restored: {name} keeps the identity the down leaves absent: its identity is not restoredre-created-empty: {name} comes back empty: its rows left with the drop |
reverse-leg-held | atomicity | note | table {table} already holds the fold the down would restore, so the reference leg leaves it |
reverse-noop | atomicity | note | left-validated: {name} is left validated: its validation contributes nothing to the downleft-not-valid: {name} is left NOT VALID by its declared policy: it contributes nothing to the down |
reverse-opaque | atomicity | note | RunRawScript {opid} reverses by its declared reverse_body, verbatim and unchecked |
reverse-recorded | atomicity | note | the range {target} is already recorded by ops {name}; what remains reaches {opid} |
reverse-slice | atomicity | note | the reverse is a slice to {opid}; {count} later operations are kept |
reverse-stop | atomicity | note | {operator} {opid} stops the reverse |
reverse-undeclared | atomicity | note | RunRawScript {opid} declares no reversal: a down over it refuses until it declares one of reverse_body, reverse_blocked or reverse_noop |
reverse-unrecorded | atomicity | note | this script is unrecorded: the oplog does not record the reverse it applies |
reverse-withdrawn | atomicity | note | DeclareReversal {opid} withdraws DeclareReversal {supersedes} for RunRawScript {target}: the script’s inline {reversal_arm} now governs the down |
reverse-writes-before-rule | atomicity | note | {operator} {opid} relaxes the rule on {name} in {table} after earlier ghost ops wrote rows |
row-axis-report | data-motion | note | {operator} moves rows of {table}: {query} |
schema-body-finding | schema | warning | {name}: {opid} {diagnostic}: {reason} |
schema-context | schema | note | {schema} |
schema-deprecated-flag | schema | warning | --as={name} is deprecated for op show; use --framing={name} and the persistent --as=text|json for output mode |
schema-drift-observed | schema | warning | added: {target} exists in the database but not in the schemaremoved: {target} exists in the schema but not in the databasemodified: {target} differs between the schema and the databaserenamed: {target} appears renamed between the schema and the databaseunmodelled: {target} carries a facet the model does not representunresolved: {target} could not be resolved against the schemadeferred: {target} was deferred to a later drift pass |
schema-emission-sentinel | schema | warning | emission-path sentinel ({name}) |
schema-fork-unmerged | schema | warning | schema {schema}: {values} share prev={opid}; the oplog carries an unmerged fork |
schema-import-drop-moved-past-reader-replace | schema | note | the drop of {target} was minted after the CREATE OR REPLACE FUNCTION at line {line} that ends its read by {reason} |
schema-import-index-key-expression-unresolved | schema | warning | index {name} in schema {schema} was set aside: its key expression could not be resolved ({reason}) |
schema-import-reader-drop-moved-ahead | schema | note | DropFunction {name} was minted ahead of the drop of {target} at line {line}, which it read; this DROP FUNCTION mints nothing |
schema-import-skip-review | schema | note | some objects were set aside during import |
schema-import-unclassified-kind | schema | warning | an object kind reached the classifier with no matching arm |
schema-init-flag-ignored | schema | warning | {name} ignored |
schema-occupancy-undeterminable | schema | warning | {reason} |
schema-op-show-replay-finding | schema | warning | replay could not classify every op in this oplog |
schema-partition-bound-timestamptz-zoneless | schema | warning | partition {name} bounds a timestamptz key by {values} with no UTC offset |
schema-partition-tree-indexed | schema | warning | partitioned table {table} holds an index, so identify and migrate --from-dump cannot place a database holding its partition tree |
schema-policy-references-unmodeled-role | schema | warning | policy {name} references role(s) {values} that retrofit does not track; verify they exist before apply |
schema-registry-empty | schema | note | no schemas are registered under {schema} |
schema-same-operator-as-tail | schema | warning | {operator} on lines {values} |
schema-stray-oplog | schema | warning | {path} |
seal-avoidable-loss | schema | note | {operator} {opid} on {name} ({element_id}) loses data that a later op of this range gives back without it |
seal-avoidable-loss-unread | schema | warning | the corpus scan failed, so this range’s avoidable losses were not read |
seed-anchor-sealed | dml | note | {table} record anchors on {opid}, inside the range seal {target} covers: a fresh build and --at render it there; a database already past seal {target} will not receive it (migrate --from {target} omits it, dml-behind-window) |
sequence-not-created-by-render | schema | warning | sequence {name} that {table}.{column} draws from is not created by the rendered DDL |
sequence-reference-unparsed | schema | warning | the DEFAULT of {table}.{column} calls nextval on a sequence the rendered DDL may not create |
snapshot-state-under-trigger | data-motion | note | the rows that state {table} load before trigger {name} exists and do not fire it; migrate --from root loads them with {name} live |
split-table-down-refused | data-motion | note | migrate --down refuses a window holding this SplitTable: the split moved rows, and its reverse is a MergeTable you author |
squash-lossy-step | schema | note | {operator} {opid} on {name} is taken out, and a database past it has already lost those values |
squash-stop | atomicity | note | {operator} {opid} stops the squash |
stale-dml-reference | dml | warning | {reason} |
table-pattern-unmatched | dml | warning | --table pattern {name} matched no statement in the source |
type-cast-breaks-dependent-call | type | warning | {target} passes {columns} to {name}, and PostgreSQL has no implicit cast from {from_type} to the parameter’s {to_type}, so the migration’s re-create of {target} after the cast fails at apply |
type-cast-session-time-zone | type | note | {target} cast from {from_type} to {to_type} with no using= converts each value in the applying session’s TimeZone |
type-creation-only | type | warning | type “{name}” is a serial pseudo-type usable only when creating a column; it is not a valid ALTER COLUMN … SET DATA TYPE target |
type-fk-partner-narrower | type | note | {target} stays {from_type} across foreign key {name} from a column cast to {to_type} |
type-unknown | type | warning | type “{name}” is not a recognized PostgreSQL built-in type |
type-varchar-without-length | type | warning | warning: type “{name}” without an explicit length is equivalent to textsuggestion: consider giving type “{name}” an explicit length |
validity-cannot-un-validate | schema | warning | {name}: the reverse of this change cannot re-validate {from_type} against {to_type} |
witness-digest-unsupported | data-motion | warning | the data record for {table} at {opid} ({path}), in the range this down reverts, has a witness whose digest algorithm is not supported |
witness-out-of-range | data-motion | warning | the data record for {table} at {opid} ({path}), in the range this down reverts, has a witness outside the seal range that covers it |