[regression](test) expand Iceberg and Paimon partition evolution coverage

[Gabriel39]

What changed

• add five Iceberg P0 suites for partition evolution with static/runtime filters, snapshot/tag/branch reads, position deletes, equality deletes, deletion vectors, and a complete format/scanner transform timeline
• add three Paimon P0 suites for fixed-partition complex schema evolution, partitioned primary-key delete/upsert/compaction, and atomic rejection of unsupported partition-key mutations
• keep independent dimensions in separate Groovy suites for parallel execution, while marking four profile-dependent suites nonConcurrent
• add deterministic runner-generated .out files and update the Iceberg/Paimon schema evolution and time-travel coverage matrix

Coverage highlights

• Iceberg add/drop/replace partition fields and identity, bucket, truncate, year/month/day/hour transforms
• complex STRUCT/MAP/ARRAY changes combined with partition filtering and historical references
• runtime-filter pruning with attributable positive profile-counter assertions on supported identity partition paths
• Parquet/ORC, file scanner V1/V2, Iceberg v2 position/equality deletes, and v3 deletion vectors
• Paimon Parquet/ORC, asserted JNI/native split paths, physical DV metadata, primary-key upsert/delete/compaction, snapshot/tag/branch
• topology-independent execution on a distributed Doris cluster

Validation

• master: f1460f89230441bc1b6b1872d66bd32a526b25b1
• one FE and two live BEs
• eight partition-evolution extension suites: 8 passed, 0 failed, 0 fatal, 0 skipped
• documented P0 matrix: 20 suites, including 8 partition-evolution extensions
• no Doris production code changed
• no additional Doris product issue was reproduced

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[Gabriel39]

run buildall

[Gabriel39]

/review

[github-actions]

Requesting changes. The patch is test/document-only and the fixtures are broad, but several matrix cells can pass without exercising the behavior they claim.

Critical checkpoint conclusions:

• Goal and proof: not yet met. The current fixtures miss post-drop position/DV data, attributable add-identity/bucket/temporal RF pruning, cross-spec temporal matches, complex-field bindings, and proven native/DV paths.
• Scope/focus: source changes are confined to seven external-table suites and one coverage document. The scope is coherent, but the document overstates several transform/delete/format/reader cross-products.
• Concurrency: no Doris product concurrency was changed. The two profile-polling suites do have a test-harness race because they run in parallel while discovering profiles through the bounded FE-wide list.
• Lifecycle/cleanup: container-process completion, catalog refresh, snapshot capture, unique fixture names, and suite-local JDBC cleanup look sound. No static-initialization or product object-lifecycle change applies.
• Configuration: no new config item was added. Session-variable changes are suite-local because SuiteContext closes each connection; the issue is whether the RF is retained and observed, not cross-suite leakage.
• Compatibility/persistence/data writes: no product protocol, storage-format, EditLog, or transaction path changed. External snapshot/tag ordering is sound, but one position/DV fixture writes its victim before the partition-field drop.
• Parallel reader/scanner paths: the explicit Iceberg V1/V2 toggles are real where present, but the full transform timelines do not cover the documented format/scanner cross-product. Paimon's second read may use native or CPP and does not prove a physical DV or a historical evolved projection.
• Test results: I did not run builds or tests, as required by the review contract. The live compile check passes; external/P0/nonConcurrent regressions are still pending. Fixed SQL results also need named qt/order_qt checks with generated .out files.
• Observability: the sole positive Iceberg RF profile is attributable to category identity pruning. The delete-aware RF checks lack plan/profile proof, and the Paimon reader/DV path lacks explain/metadata proof.
• Performance: no product hot path changed. The main execution concern is nondeterministic profile retention under parallel suite traffic.
• User focus: no additional user-provided focus was supplied.

The PR title checker is also failing because test: expand Iceberg and Paimon partition evolution coverage does not match the required [type](module) form. Suggested title: [test](regression) Expand Iceberg and Paimon partition evolution coverage.

[Gabriel39]

run buildall

[Gabriel39]

/review

[github-actions]

Codex automated review failed and did not complete.

Error: You've hit your usage limit. Visit https://chatgpt.com/codex/settings/usage to purchase more credits or try again at Jul 30th, 2026 7:58 AM.
Workflow run: https://github.com/apache/doris/actions/runs/30081551337

Please inspect the workflow logs and rerun the review after the underlying issue is resolved.

[github-actions]

PR approved by at least one committer and no changes requested.

1