Open-source manufacturing ERP, MES and QMS. Quoting, MRP, inventory, shop floor, quality and lot/serial traceability on one Postgres schema, with a REST API and MCP server. Self-host or use Carbon Cloud.
TypeScript ai-agentsbill-of-materialserpinventory-managementleanmanufacturingmanufacturing-execution-systemmcp-servermesmrpopen-source-erppostgresqlproduction-planningqmsquality-management-systemreact-routerself-hostedsupabasetraceabilitytypescript
fix(production): a serial job with its units assigned completes from the desk (#1701)
Anshul sharma8213240df66d
diff --git a/packages/database/supabase/migrations/20260922183000_serial-job-desk-completion.sql b/packages/database/supabase/migrations/20260922183000_serial-job-desk-completion.sqlnew file mode 100644index 0000000000..850212bec7--- /dev/null+++ b/packages/database/supabase/migrations/20260922183000_serial-job-desk-completion.sql@@ -0,0 +1,102 @@+-- A serial job whose units already exist completes from the desk again.+--+-- 20260922050131 refused EVERY serial-tracked job that had an operation+-- shortfall, on the assumption that a missing operation quantity meant a+-- missing serial unit. It does not. Serial numbers are assigned at job+-- CREATION (assignJobSerialNumbers, on the create paths only), so the ordinary+-- desk flow -- create a serial job, release it, never open the Operations tab,+-- press Complete -- has every numbered unit Reserved and ready to receive while+-- each operation still reads 0. Those jobs completed before that migration and+-- were refused after it, with "a serial number is required" naming serials that+-- were sitting right there.+--+-- Fix: drop the guard. complete_job_to_inventory's serial branch already+-- refuses a job genuinely short of units ("has N serial unit(s) left to+-- receive, fewer than the M being completed") after trying the un-numbered seed+-- fallback, and it runs in the same transaction, so a real shortage still rolls+-- the completion back. The guard could only ever duplicate that decision from+-- worse evidence: operation quantities rather than the tracked entities the+-- receipt actually consumes. Deleting it removes the disagreement instead of+-- teaching two places to agree.+--+-- Recreated VERBATIM from 20260922050131_mark-complete-completes-remaining-quantities.sql+-- minus the serial guard and its now-unused v_serial_* locals.++CREATE OR REPLACE FUNCTION complete_job_remaining_quantities(+ p_job_id TEXT,+ p_quantity_complete NUMERIC,+ p_user_id TEXT+)+RETURNS VOID+LANGUAGE plpgsql+SECURITY DEFINER+SET search_path = public+AS $$+DECLARE+ v_shortfall NUMERIC;+BEGIN+ -- Only the manual path closes operations. Under a trigger this is the MES+ -- path, where the floor reports its own quantities — and where writing a+ -- sibling jobOperation row fails with "tuple to be updated was already+ -- modified by an operation triggered by the current command", since+ -- sync_finish_job_operation is a BEFORE trigger mid-statement on those rows.+ IF pg_trigger_depth() > 0 THEN+ RETURN;+ END IF;++ -- The largest shortfall across the terminal operations. MAX, not SUM: these+ -- are parallel paths to the same finished part and each must report the full+ -- quantity, so summing them would invent units nobody required. Reworked+ -- units already satisfy the auto-Done predicate, so they count as reported.+ SELECT COALESCE(MAX(+ p_quantity_complete+ - COALESCE(t."quantityComplete", 0)+ - COALESCE(t."quantityReworked", 0)+ ), 0)+ INTO v_shortfall+ FROM terminal_job_operations(p_job_id) t;++ IF v_shortfall <= 0 THEN+ RETURN;+ END IF;++ -- Status as well as quantity. The auto-Done predicate lives in+ -- sync_update_job_operation_quantities, which fires on productionQuantity+ -- writes — not on a jobOperation update — so nothing else would close these+ -- operations, and a quantity-only write would leave the routing open behind a+ -- Completed job: the very symptom this function exists to remove.+ --+ -- Setting status here fires sync_finish_job_operation on each row. The caller+ -- must therefore have marked the job Completed already: the interceptor's+ -- status gate then returns early. Called while the job is still open it would+ -- instead re-enter complete_job_to_inventory and, being a BEFORE row trigger+ -- reading the pre-update quantityComplete, receive the goods a second time.+ --+ -- Every open terminal operation is closed, not only the short ones: an+ -- operation whose quantity already meets the target is Done by the same+ -- predicate the interceptor uses, so leaving it Ready on a Completed job+ -- would recreate the inconsistency this function exists to remove.+ --+ -- Reworked units already count as reported, so the new quantityComplete is+ -- the target MINUS them. Assigning the full target on top of a reworked unit+ -- would report more units than the operation ever produced (target 3 with 1+ -- complete + 1 reworked must end at 2 complete + 1 reworked, not 3 + 1).+ -- GREATEST never lowers a quantity the floor already reported.+ UPDATE "jobOperation" jo+ SET "quantityComplete" = GREATEST(+ COALESCE(t."quantityComplete", 0),+ p_quantity_complete - COALESCE(t."quantityReworked", 0)+ ),+ status = 'Done',+ "updatedBy" = COALESCE(p_user_id, jo."updatedBy", jo."createdBy"),+ "updatedAt" = NOW()+ FROM terminal_job_operations(p_job_id) t+ WHERE jo.id = t.id;+END;+$$;++-- SECURITY DEFINER with no p_company_id to bind the call to a tenant; it stays+-- an internal helper of complete_job_to_inventory, which does enforce the+-- company check. CREATE OR REPLACE preserves grants, but re-assert them so the+-- revoke cannot be lost if this function is ever recreated from this file.+REVOKE ALL ON FUNCTION complete_job_remaining_quantities(TEXT, NUMERIC, TEXT) FROM PUBLIC, anon, authenticated;diff --git a/packages/database/supabase/tests/job-completion-operation-quantities.test.sql b/packages/database/supabase/tests/job-completion-operation-quantities.test.sqlindex a15eb03d13..7d792d7c85 100644--- a/packages/database/supabase/tests/job-completion-operation-quantities.test.sql+++ b/packages/database/supabase/tests/job-completion-operation-quantities.test.sql@@ -218,19 +218,56 @@ BEGINASSERT v_row.status <> 'Done' AND v_row."quantityComplete" = 0,'A sub-assembly operation must not be completed by the job: ' || v_row.status || ' @ ' || v_row."quantityComplete";- -- Serial job with units still outstanding is refused, and nothing is applied.+ -- REGRESSION: a serial job whose units already exist completes from the desk.+ --+ -- Serial numbers are assigned at job CREATION, so the ordinary desk flow --+ -- create, release, never open the Operations tab, press Complete -- has every+ -- numbered unit Reserved and ready while the operations still read 0. The+ -- first version of this guard refused exactly that, naming a serial number+ -- that was sitting right there. Operation shortfall is not evidence of a+ -- missing serial; only the absence of un-received units is.v_job := pg_temp.make_job(v_company_id, v_location_id, v_serial_item, v_part, 'SJ', 3);v_op1 := pg_temp.add_operation(v_job, v_company_id, v_process, v_work_center, 3, 1);v_error := pg_temp.try_complete(v_job, 3);- ASSERT v_error LIKE '%serial number is required%',- 'Serial job with outstanding units was not refused: ' || COALESCE(v_error, '<succeeded>');+ ASSERT v_error IS NULL,+ 'Serial job with its units already assigned was refused: ' || COALESCE(v_error, '');+ SELECT status, "quantityComplete" INTO v_row FROM job WHERE id = v_job;+ ASSERT v_row.status = 'Completed' AND v_row."quantityComplete" = 3,+ 'Serial job with assigned units did not complete from the desk';+ ASSERT pg_temp.ops(v_job) = 'Done:3',+ 'Serial desk completion left its operations open: ' || pg_temp.ops(v_job);+ -- Serial receipts post one itemLedger row per numbered unit, so three units+ -- are three rows totalling three -- not a single row of three.+ ASSERT pg_temp.receipts(v_job) = '3:3',+ 'Serial desk completion did not receive its three units: ' || pg_temp.receipts(v_job);++ -- A serial job genuinely short of numbered units is still refused, and+ -- nothing is applied. Here the units are consumed elsewhere, so there is+ -- nothing left for this job to receive.+ --+ -- The refusal comes from complete_job_to_inventory's serial branch, not from+ -- complete_job_remaining_quantities: the receipt is the code that knows which+ -- tracked entities it can actually consume, so it is the only place that can+ -- tell a numbered-but-unreported job from a genuinely short one.+ v_job := pg_temp.make_job(v_company_id, v_location_id, v_serial_item, v_part, 'SJ-SHORT', 3);+ v_op1 := pg_temp.add_operation(v_job, v_company_id, v_process, v_work_center, 3, 1);+ UPDATE "trackedEntity" SET status = 'Consumed'+ WHERE attributes->>'Job Make Method' = (+ SELECT id FROM "jobMakeMethod" WHERE "jobId" = v_job AND "parentMaterialId" IS NULL+ );++ v_error := pg_temp.try_complete(v_job, 3);+ ASSERT v_error LIKE '%serial unit(s) left to receive%',+ 'Serial job with no available units was not refused: ' || COALESCE(v_error, '<succeeded>');SELECT status INTO v_row FROM job WHERE id = v_job;ASSERT v_row.status <> 'Completed', 'A refused serial completion still completed the job';ASSERT pg_temp.ops(v_job) = 'Ready:0','A refused serial completion still wrote quantities: ' || pg_temp.ops(v_job);-- The same serial job completes once the floor has reported every unit.+ v_job := pg_temp.make_job(v_company_id, v_location_id, v_serial_item, v_part, 'SJ-MES', 3);+ v_op1 := pg_temp.add_operation(v_job, v_company_id, v_process, v_work_center, 3, 1);UPDATE "jobOperation" SET "quantityComplete" = 3 WHERE id = v_op1;v_error := pg_temp.try_complete(v_job, 3);ASSERT v_error IS NULL, 'Serial job finished in MES failed to complete: ' || COALESCE(v_error, '');diff --git a/packages/jobs/manifests/schema.json b/packages/jobs/manifests/schema.jsonindex 13ba2d609c..08fba19476 100644--- a/packages/jobs/manifests/schema.json+++ b/packages/jobs/manifests/schema.json@@ -1,11 +1,11 @@{"kind": "carbon-company-backup","version": 1,- "schemaVersion": "20260922050131",+ "schemaVersion": "20260922183000","sourceCompanyId": "schema-baseline","sourceCompanyGroupId": null,"sourceCompanyName": null,- "exportedAt": "2026-09-22T07:16:49.167Z",+ "exportedAt": "2026-09-22T10:59:26.429Z","exportedBy": "schema-baseline","label": "schema baseline","includeStorage": "none",