Known limitations (2.0)

  • Several systems can be connected, but the views show one (the active) system at a time.
  • Where used builds its cross-reference with DSPPGMREF in QTEMP, so it needs a Mapepire SQL engine (not db2util). Scanning large libraries takes a while.
  • Source dates and conflict checks need a Mapepire SQL engine. With db2util, members are copied as plain text and dates are reset on save.
  • Long lines: lines longer than the source file’s record length can’t be saved; the save stops and the previous version is kept.
  • Table data editor writes each change straight away (no commitment control). Rows are identified by relative record number, so don’t reorganize a file while editing it.
  • Rename only works inside one source and refuses length changes in fixed-format sources, because they would shift columns.
  • Fixed → free converter converts H, F, D, P and C specs; I and O specs and primary/table files are left with a // TODO.
  • Generate SQL (DDL) needs a Mapepire SQL engine.
  • Extract to procedure works on whole lines of free-form calculations and creates a procedure without parameters.
  • Library compare decides “different” from line counts, change dates, sizes and source timestamps; it does not read object contents.
  • Table data compare needs a Mapepire SQL engine; it lists at most 200 rows per side and 500 changed rows (the counts are exact) and skips LOB/XML columns.
  • Journal explorer shows at most 500 entries and the record images as text (CCSID 37), so packed and binary fields look garbled; before images of updates need IMAGES(*BOTH).
  • Authority checker follows the usual IBM i order (*ALLOBJ, own, groups, authorization list, *PUBLIC) but does not consider adopted authority or field-level authority.
  • PTF & system health: the comparison with IBM’s PTF levels needs internet access from the IBM i; certificate and default-password checks need ALLOBJ and *SECADM. Checks that can’t run are listed as *not checked.
  • Data queues are peeked as text; DDM data queues can’t be peeked.
  • Live alerts poll the IBM i (every 60 s by default) only while VS Code is open and connected; they are off until you turn them on.
  • Object promotion copies with SAVOBJ / RSTOBJ: objects in use in the target can’t be replaced, and Undo restores the replaced objects but does not delete objects that were new.
  • Storage advisor: “last used” dates are kept by IBM i but not for every kind of use (for example some SQL reads) — check before deleting.
  • Import reads the first worksheet of an .xlsx file and plain CSV; formulas are imported as their last calculated value, and rows are inserted without commitment control.
  • RPGUnit needs iRPGUnit installed on the IBM i; per-test results need a version of RUCALLTST with XMLSTMF, otherwise only the overall result is shown.
  • Code metrics and documentation come from the source text only (no compile listing), so they are estimates.
  • Protection decides from the command or SQL statement itself; a program you CALL can still change things inside it, so Production asks before every CALL and Read-only refuses it.
  • Screen preview draws every field as if all indicators were on, and shows reference fields (REFFLD) without a length as 10 long; it does not run the program.
  • Database diagram: inferred relationships are guesses from column names — check them before relying on them.
  • Spool to PDF (local) uses the spool’s text, so fonts, barcodes and graphics of AFP spools are not kept; use vanthrex.spool.pdfMode: server for those. E-mail needs SMTP set up on the IBM i.
  • Print preview prints each record format once and shows every field as if all indicators were on; it does not run the program.
  • Find and replace reads at most 1,000 members of a source file and skips members someone has open.
  • Test data fills every column with made-up values that fit its type; constraints, triggers and check rules on the table may still refuse some rows.
  • Debugger needs the IBM i Debug Service and IBM’s IBM i Debug extension, and debugs programs in a batch job (no service entry points or 5250 screens yet) — see Debugging.
  • AI assistant needs a chat model in VS Code (for example GitHub Copilot Chat). The code you ask about is sent to that model. Its read-only check blocks known side effects, but user-defined SQL functions can do anything — that is why Vanthrex asks before every query.
  • Call graph and SQL Explain need a Mapepire SQL engine. The call graph can’t see dynamic calls (program names in variables).
  • Git sync compares member change times and text; it does not track renamed or deleted members (deleted members are reported and kept in the repository).
  • Interactive (5250) commands such as WRKACTJOB can’t run from the CL runner; use their OUTPUT(*PRINT) form or the IBM i Services SQL snippets (type ibmi- in a .sql file).

Have an idea or hit a limit that blocks you? Open an issue.