[enjoykumawat]

Summary

• Bug: When multiple tool calls run in parallel and each writes to the same state_delta key containing a list value, deep_merge_dicts silently drops all but the last value because lists hit the else branch and get overwritten.
• Fix: Add a list-type check in deep_merge_dicts so list values are concatenated (d1[key] + value) instead of overwritten, preserving all entries from parallel function responses.
• Tests: Added 5 unit tests covering list concatenation, scalar overwrite, nested dict merge, mixed-type handling, and an integration test for merge_parallel_function_response_events with state_delta lists.

Reproduction

# Tool A's delta: {"state_delta": {"items": ["a"]}}
# Tool B's delta: {"state_delta": {"items": ["b"]}}
# Before fix: {"state_delta": {"items": ["b"]}} — item "a" is lost
# After fix:  {"state_delta": {"items": ["a", "b"]}} — both preserved

Test plan

• All 5 new unit tests pass (test_deep_merge_dicts_* and test_merge_parallel_function_response_events_merges_state_delta_lists)
• All 32 existing tests in test_functions_simple.py pass (0 regressions)
• Manual: create parallel tool calls that accumulate list state and verify no data loss

Fixes #5190

[wuliang229]

Please move the tests to test_functions_parallel.py

[enjoykumawat]

Done! Moved all the deep_merge_dicts and state_delta list tests to test_functions_parallel.py in commit 4b3db55. Also merged latest main to pick up any upstream changes. CI is re-running now.

[enjoykumawat]

Hi @wuliang229, just following up — the tests have been moved to test_functions_parallel.py as requested (commit 4b3db55) and CI is passing. Would you have a chance to re-review when you get a moment? Happy to make any additional changes if needed.

[adk-bot]

Thank you @enjoykumawat for your contribution! 🎉

Your changes have been successfully imported and merged via Copybara in commit 1ff84eb.

Closing this PR as the changes are now in the main branch.

[enjoykumawat]

Hi @GWeale — I see this was reopened after the Copybara merge (commit 1ff84eb is already in main). Is there a follow-up you'd like from me, or was this reopened for internal tracking? Happy to help.

[enjoykumawat]

Quick question before this gets another look — digging into the Copybara history here, I see this was actually merged internally (1ff84eb, 2026-06-22) and then reverted about 4 hours later by an internal commit (fda23474, same day), with no reason given in either commit message. The PR was then reopened.

Was there a concern found internally with this approach (the list-concatenation logic in deep_merge_dicts, or something else) that led to the revert? I've since rebased this onto current main — the bug is still present there (list values in state_delta still get silently overwritten instead of concatenated) — and all tests pass, but I'd rather understand what went wrong the first time than have the same thing happen again. Happy to adjust the approach if there's a reason this shouldn't land as-is.

[GWeale]

thanks for this, and sorry your question about the revert went unanswered. concatenating lists here broke existing callers each time it landed, so main now keeps the latest write when parallel calls update the same key instead: d42a3be. closing