Problem
A file defining the same test function twice is caught sequentially and ignored
entirely under --parallel — the mode CI runs.
function test_same() { assert_same 1 1; }
function test_same() { assert_same 2 2; }$ bashunit --no-parallel dup_test.sh
Duplicate test functions found
Duplicate functions: test_same # exit 1
$ bashunit --parallel dup_test.sh
✓ Passed: Same
All tests passed # exit 0
This matters because bash silently discards the first definition: one of the two
tests never runs. The check exists precisely to catch that, and it is off in the
mode most projects use in CI.
Cause
check_duplicate_functions runs inside call_test_functions, which
load_test_files backgrounds per file under --parallel. The state it sets
dies with the subshell, exactly like the counter in #1145 — only .result
files cross that boundary.
Fix
Run the duplicate check in the parent, before the per-file dispatch, so it is
independent of the run mode.
Also
The report names no line numbers:
Duplicate functions: test_same
In a long test file that means hunting for the second definition by hand. The
awk pass already walks every line, so recording them costs nothing.
Problem
A file defining the same test function twice is caught sequentially and ignored
entirely under
--parallel— the mode CI runs.This matters because bash silently discards the first definition: one of the two
tests never runs. The check exists precisely to catch that, and it is off in the
mode most projects use in CI.
Cause
check_duplicate_functionsruns insidecall_test_functions, whichload_test_filesbackgrounds per file under--parallel. The state it setsdies with the subshell, exactly like the counter in #1145 — only
.resultfiles cross that boundary.
Fix
Run the duplicate check in the parent, before the per-file dispatch, so it is
independent of the run mode.
Also
The report names no line numbers:
In a long test file that means hunting for the second definition by hand. The
awk pass already walks every line, so recording them costs nothing.