Skip to content

fix(ci): wire stage matrix size into runtime#19606

Merged
sundy-li merged 3 commits intodatabendlabs:mainfrom
sundy-li:issue-19598-20260324-2020
Apr 17, 2026
Merged

fix(ci): wire stage matrix size into runtime#19606
sundy-li merged 3 commits intodatabendlabs:mainfrom
sundy-li:issue-19598-20260324-2020

Conversation

@sundy-li
Copy link
Copy Markdown
Member

@sundy-li sundy-li commented Mar 24, 2026

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

Fixes #19598

  • pass the stage matrix size dimension into the sqllogic stage action
  • export TEST_STAGE_SIZE so prepare_stage.sh can switch between small and large
  • add a lightweight regression check to make lint-yaml for this wiring

Changes

  • add a required size input to .github/actions/test_sqllogic_stage
  • pass ${{ matrix.size }} from .github/workflows/reuse.sqllogic.yml
  • add .github/scripts/check_sqllogic_stage_size_wiring.py and call it from lint-yaml

Tests

  • Regression check script: python3 .github/scripts/check_sqllogic_stage_size_wiring.py
  • YAML parse validation for the touched workflow and action files

Type of change

  • Bug fix (non-breaking change which fixes an issue)

This change is Reviewable

@github-actions github-actions Bot added the pr-bugfix this PR patches a bug in codebase label Mar 24, 2026
@sundy-li sundy-li added agent-reviewable Ready for agent review agent-approved Approved by agent and removed agent-reviewable Ready for agent review labels Mar 24, 2026
Comment thread Makefile Outdated
@sundy-li sundy-li merged commit 9a5c512 into databendlabs:main Apr 17, 2026
9 checks passed
@sundy-li sundy-li deleted the issue-19598-20260324-2020 branch April 17, 2026 00:42
@sundy-li
Copy link
Copy Markdown
Member Author

I tracked the stage parquet failure to row-group start row handling, not the tests.

Root cause:

  • TEST_STAGE_SIZE=large makes prepare_stage.sh set parquet_fast_read_bytes=0
  • that switches stage parquet reads to the row-group path
  • in src/query/storages/parquet/src/source.rs, start_row was incremented before building the current ParquetRowGroupPart
  • so metadata$file_row_number used the end offset of the current row group instead of its start

Fix pushed in this PR:

  • fix: correct parquet row-group start row
  • also fixed the same pre-increment pattern in src/query/storages/parquet/src/parquet_variant_table/source.rs

Local check:

  • cargo test -p databend-common-storages-parquet --no-run

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

agent-approved Approved by agent pr-bugfix this PR patches a bug in codebase

Projects

None yet

Development

Successfully merging this pull request may close these issues.

bug: stage matrix size dimension is not wired into runtime

2 participants