validate column structure before applying patches

[seva-potapov]

Changelog category (leave one):

• Bug Fix (user-visible misbehavior in an official stable release)

Changelog entry (a user-readable short description of the changes that goes into CHANGELOG.md):

validate column structure before applying patches

Details:

applyPatchesToBlockRaw matches patch columns to result columns by name only, without type validation. When schema evolution changes a column type (e.g., UInt32 change to String), the patch column has a different type than the result column that results in crash.

[clickhouse-gh]

Workflow [PR], commit [2c2c2d3]

Summary: ✅

---

AI Review

Summary

This PR hardens patch application across schema evolution by validating per-column types before applying updates, adds per-source filtering in the combined patch path, and adds regression coverage for raw and mixed-source scenarios. I did not find additional high-confidence correctness/safety/performance issues in the current PR head beyond already-posted inline discussion.

Missing context

• ⚠️ No CI run/log context was provided in this request, so this review is code-only.

ClickHouse Rules

Item	Status	Notes
Deletion logging	➖
Serialization versioning	➖
Core-area scrutiny	✅
No test removal	✅
Experimental gate	➖
No magic constants	✅
Backward compatibility	✅
SettingsHistory.cpp	➖
Safe rollout	✅
Compilation time	✅

Final Verdict

• Status: ✅ Approve

[clickhouse-gh]

LLVM Coverage Report

Metric	Baseline	Current	Δ
Lines	83.70%	83.70%	+0.00%
Functions	23.90%	23.90%	+0.00%
Branches	76.20%	76.20%	+0.00%

PR changed lines: PR changed-lines coverage: 98.41% (186/189, 1 noise lines excluded)
Diff coverage report
Uncovered code

[alexey-milovidov]

Is it possible to have an end-to-end .sql/.sh test in addition to a unit test?

[seva-potapov]

I have looked into it, Alexey, but to replicate the issue we've had on the cloud, we need SMT and shared catalog in order to arrive at schema evolution outside of ALTER MODIFY (which doesn't have the problem we are fixing here).

[evillique]

I think this fix is incorrect; we silently lose the data from LWU that was in progress if we ALTER the column type.

This fix looks more correct to me: #100107

[seva-potapov]

thank you, @evillique ! your fix is definitely more correct.