test gms type fallback

[jycor]

No description provided.

[jycor]

#benchmark

[github-actions]

	Main	PR
covering_index_scan_postgres	2111.28/s	2140.45/s	+1.3%
groupby_scan_postgres	103.73/s	${\color{lightgreen}138.30/s}$	${\color{lightgreen}+33.3\%}$
index_join_postgres	649.41/s	704.67/s	+8.5%
index_join_scan_postgres	859.34/s	883.20/s	+2.7%
index_scan_postgres	34.23/s	33.34/s	-2.7%
oltp_delete_insert_postgres	863.42/s	845.47/s	-2.1%
oltp_insert	717.12/s	731.09/s	+1.9%
oltp_point_select	3549.74/s	3622.43/s	+2.0%
oltp_read_only	3456.92/s	3512.36/s	+1.6%
oltp_read_write	2534.40/s	2577.01/s	+1.6%
oltp_update_index	795.35/s	795.31/s	-0.1%
oltp_update_non_index	855.34/s	849.98/s	-0.7%
oltp_write_only	1855.22/s	1811.96/s	-2.4%
select_random_points	2093.14/s	2098.92/s	+0.2%
select_random_ranges	1611.87/s	1636.69/s	+1.5%
table_scan_postgres	32.49/s	33.13/s	+1.9%
types_delete_insert_postgres	817.91/s	832.01/s	+1.7%
types_table_scan_postgres	14.91/s	14.39/s	-3.5%

[itoqa]

Commit: 6fe4620: 11 test cases ran, 11 passed ✅.

Summary

Coverage spans core database behavior for creating, changing, querying, and removing data and schemas; type identity and value conversion across numeric, text, date, JSON, and geometry-related values; concurrent catalog consistency; connection recovery; and clean startup, shutdown, reset, and isolation flows. It includes happy paths plus edge cases around unsupported values, nulls, failed conversion attempts, cancellation, and reuse of the service afterward.

Safe to merge — the exercised behaviors completed successfully with no PR-attributable regressions, new failures, or previously reported failures remaining. No merge blocker was identified; the unsupported-value scenarios were handled as controlled errors with the service remaining usable.

Tests run by Ito

Result	Severity	Type	Description
✅	—	Catalog	A created catalog_items table exposed its catalog_items composite row type and _catalog_items implicit array type with stable OIDs. Row and array composite expressions returned the inserted 7/alpha/12.50 values successfully.
✅	—	Catalog	Two concurrent sessions and a fresh session exposed the same composite and implicit-array types with matching names, OIDs, and definitions after schema changes.
✅	—	Conversion	Local Doltgres returned the expected numeric, text, and date values, preserved PostgreSQL type metadata, and evaluated derived expressions, filtering, ordering, and aggregates correctly.
✅	—	Conversion	Supported numeric and date expressions, NULL handling, and subsequent recovery queries returned the expected values and PostgreSQL types after candidate unsupported conversion attempts. The required non-null unsupported GMS value could not be constructed because available geometry syntax was rejected before reaching the conversion boundary, while source inspection confirms conversion errors are isolated to the individual call.
✅	—	Engine	The upgraded engine built successfully, accepted connections, completed SELECT plus CREATE/INSERT/SELECT/DROP table operations, and shut down without a lingering server process.
✅	—	Engine	The server-owning and shared-server worker databases kept distinct marker rows, and resetting either database recreated an empty state without cross-contamination.
✅	—	Engine	The fresh database accepted CREATE TYPE, CREATE TABLE, ALTER TABLE, INSERT, and SELECT, and dropping the objects removed them from the catalog.
✅	—	Engine	SIGTERM terminated the running Doltgres server, and a fresh initialization on the temporary data directory became ready and successfully returned SELECT 1 and catalog data.
✅	—	Fallback	Local Doltgres returned boolean, smallint, integer, and bigint metadata with the expected values, and the related mapping package tests passed.
✅	—	Fallback	The local Doltgres server returned the expected PostgreSQL type metadata and usable values for floating-point, decimal, temporal, text, and JSON categories, and the mapping package tests passed.
✅	—	Fallback	Unsupported geometry conversion returned the documented controlled error, the mapping package tests passed, and a subsequent SELECT 1 succeeded on the local Doltgres session.

Tip

Reply with @itoqa to send us feedback on this test run.

[itoqa]

Ito Diff Report — 6fe4620 → 0bf8bd0: 4 test cases ran, 4 passing ✅.

Diff Summary

The run covered core database behavior across startup and shutdown, schema and catalog changes, data type conversion, composite values, filtering, ordering, functions, aggregates, persistence, recovery, and safe handling of unsupported values. It exercised normal workflows as well as edge and error-recovery paths, with healthy behavior throughout the areas checked.

Safe to merge — the exercised behavior showed no regressions, new failures, or previously flagged failures attributable to this PR. No merge-blocking issues were found; the overall result is low risk.

Tests run by Ito

Result	State	Severity	Type	Description
✅	Passing	—	Catalog	Created catalog_probe and confirmed its catalog composite row type and implicit _catalog_probe array type with OIDs. Row and array composite expressions returned the inserted values and supported field access without type-resolution errors.
✅	Passing	—	Conversion	Local Doltgres correctly converted numeric, text, date, and timestamp values across projections, filtering, ordering, functions, aggregates, and a follow-up query; PostgreSQL type metadata matched the expected mapped types.
✅	Passing	—	Engine	The locally built engine started successfully, accepted SELECT, DDL, INSERT, and readback queries, and then shut down cleanly after SIGTERM without corrupting its data directory.
✅	Passing	—	Fallback	A fresh local Doltgres session returned the expected PostgreSQL type metadata and values for Boolean, numeric, temporal, text, JSON, NULL, unknown, and record fallback cases; a follow-up query on the same connection also succeeded.
⏸️	Skipped	—	Catalog	Two concurrent sessions and a fresh session exposed the same composite and implicit-array types with matching names, OIDs, and definitions after schema changes.
⏸️	Skipped	—	Conversion	Supported numeric and date expressions, NULL handling, and subsequent recovery queries returned the expected values and PostgreSQL types after candidate unsupported conversion attempts. The required non-null unsupported GMS value could not be constructed because available geometry syntax was rejected before reaching the conversion boundary, while source inspection confirms conversion errors are isolated to the individual call.
⏸️	Skipped	—	Engine	The server-owning and shared-server worker databases kept distinct marker rows, and resetting either database recreated an empty state without cross-contamination.
⏸️	Skipped	—	Engine	The fresh database accepted CREATE TYPE, CREATE TABLE, ALTER TABLE, INSERT, and SELECT, and dropping the objects removed them from the catalog.
⏸️	Skipped	—	Engine	SIGTERM terminated the running Doltgres server, and a fresh initialization on the temporary data directory became ready and successfully returned SELECT 1 and catalog data.
⏸️	Skipped	—	Fallback	Local Doltgres returned boolean, smallint, integer, and bigint metadata with the expected values, and the related mapping package tests passed.
⏸️	Skipped	—	Fallback	The local Doltgres server returned the expected PostgreSQL type metadata and usable values for floating-point, decimal, temporal, text, and JSON categories, and the mapping package tests passed.
⏸️	Skipped	—	Fallback	Unsupported geometry conversion returned the documented controlled error, the mapping package tests passed, and a subsequent SELECT 1 succeeded on the local Doltgres session.

Tip

Reply with @itoqa to send us feedback on this test run.

[github-actions]

	Main	PR
Total	42090	42090
Successful	18459	18460
Failures	23631	23630
Partial Successes1	5284	5284

	Main	PR
Successful	43.8560%	43.8584%
Failures	56.1440%	56.1416%

${\color{lightgreen}Progressions (1)}$

copyselect

QUERY: drop table test3;

Footnotes

1. These are tests that we're marking as Successful, however they do not match the expected output in some way. This is due to small differences, such as different wording on the error messages, or the column names being incorrect while the data itself is correct. ↩