GH-50219: [R] Fix duckdb test for dbplyr 2.6.0

[thisisnic]

Rationale for this change

dbplyr 2.6.0 changed remote_con() to read from x$con instead of x$src$con. The duckdb test that simulates a non-DuckDB connection was overriding the old path, so the expected error was no longer thrown.

What changes are included in this PR?

Update the test to override ds_rt$con instead of ds_rt$src$con.

Are these changes tested?

This is a test fix — the test itself verifies the behavior.

Are there any user-facing changes?

No.

• GitHub Issue: [R] Test failure with dbplyr 2.6.0 - remote_con() path changed from x$src$con to x$con #50219

[thisisnic]

@github-actions crossbow submit test-r-depsource-system test-r-gcc-11 test-r-gcc-12 test-r-m1-san test-r-offline-maximal test-r-versions

[Copilot]

Pull request overview

Fixes an R test failure caused by dbplyr 2.6.0 changing how remote_con() retrieves the connection from a tbl.

Changes:

• Update the DuckDB integration test to override ds_rt$con (instead of ds_rt$src$con) so the simulated “non-DuckDB connection” path still triggers the expected to_arrow() error with dbplyr 2.6.0.

[github-actions]

Revision: a468619

Submitted crossbow builds: ursacomputing/crossbow @ actions-74af3d833c

Task	Status
test-r-depsource-system
test-r-gcc-11
test-r-gcc-12
test-r-m1-san
test-r-offline-maximal
test-r-versions

[thisisnic]

Needs skipping or fails on CI with cached dbplyr with old syntax

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

[raulcd]

not an expert but this looks reasonable and fixes CI, @jonkeane do you want to take a quick look before I merge?

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit d6bba6b.

There were 2 benchmark results indicating a performance regression:

• Commit Run on test-mac-arm at 2026-06-22 10:30:02Z
  • tpch (R) with engine=arrow, format=native, language=R, memory_map=False, query_id=TPCH-02, scale_factor=1
  • tpch (R) with engine=arrow, format=native, language=R, memory_map=False, query_id=TPCH-11, scale_factor=1

The full Conbench report has more details. It also includes information about 216 possible false positives for unstable benchmarks that are known to sometimes produce them.