Changelog

Trekport release notes.

The engineering changes behind the Oracle-to-PostgreSQL migration platform. Each entry is real work that landed in the repository, not roadmap.

Reserve priority accessRead the docs

Built by the team running the migration.

v0.9

Audit-grade posture across the cutover surface.

Cutover is what everyone watches. This release tightens the seams between extraction, conversion, deployment, and validation so the audit trail tells one consistent story end to end.

Improvements

  • Per-object diagnostics: every failure is bucketed into converter / catalog / extension / out-of-scope before it leaves the engine.
  • Cutover scorecard wired into the desktop with confidence-weighted overall, six validation strategies, and per-strategy collapsible drilldowns.
  • Validation report viewer renders ROW_COUNT, CHECKSUM, SAMPLE_DIFF, SEMANTIC, SCHEMA_MATCH, and EXECUTION as discrete sections.
v0.8

Fourteen-phase deployer reaches convergence.

The deployment orchestrator now drives the full phased install — roles and privileges through synonyms and grants — with iterative convergence and dependency-aware retry inside each phase.

Improvements

  • Iterative six-pass convergence per phase resolves cross-object dependencies without operator nudging.
  • Package SPEC and BODY are merged at convert time so PL/SQL bodies deploy as a single logical unit.
  • Type topological sort orders composite type creation correctly when types depend on other types.
  • Per-file conversion timeout (30s DDL, 60s PL/SQL) keeps pathological objects from stalling the pipeline.
v0.7

Three distributions, one engine.

The same engine now backs the desktop, the CLI, and the in-process facade. The desktop's deployment phases route through the engine's orchestrator instead of carrying a parallel implementation.

Improvements

  • Desktop deploy collapses into a single phase that calls the engine's orchestrator through the GraalPy facade.
  • Checkpoint resumption with a real banner on the wizard: cancel mid-deploy, reopen, pick up where the run stopped.
  • Auto-retry with exponential backoff on transient deployment errors; only transient classes are retried.
  • Eighty-plus diagnostic call sites moved from eprintln onto structured log levels.
v0.6

Catalog extractor parity with the CLI.

The desktop's Rust catalog extractor now produces the same directory layout the CLI does. The Phase 0 _database tree carries everything an Oracle target depends on before any schema objects are created.

Improvements

  • Phase 0 _database tree: profiles, system privileges, role privileges, contexts, public synonyms, scheduler jobs.
  • Dependency graph with Kahn's topological sort so type creation order matches Oracle's resolution order.
  • Per-schema policies, security, and queues directories for VPD, redaction, TDE encrypted columns, and AQ queue tables.
  • Invalid objects audit emitted alongside the extracted catalog so reviewers see Oracle-side gaps before deploy starts.
v0.5

Compatibility layer auto-installs.

An Oracle migration succeeds or fails on the gap between dialects. Trekport now installs the compatibility extensions the engine needs in Phase 0, before any schema object has a chance to fail on a missing function.

Improvements

  • orafce, pgcrypto, pg_trgm, btree_gist, pg_cron, http, and file_fdw install automatically when the target permits.
  • Compatibility step inserted into the migration wizard so the operator confirms what gets installed before deploy runs.
  • Engine prefers official PostgreSQL extensions over custom shims and only ships proprietary fills when no community equivalent exists.
  • Extension load is idempotent — a re-run on a clean target completes with zero diff.

More release notes land as engineering wraps each wave.

Teams on the waitlist receive the engineering notes directly when their wave opens.

Reserve priority access