Reset string array size when allocation fails in init - #593

Description

rcutils_string_array_init() writes the requested size into the array before it tries to allocate:

string_array->size = size;
string_array->data = allocator->zero_allocate(size, sizeof(char *), allocator->state);
if (NULL == string_array->data && 0 != size) {
  RCUTILS_SET_ERROR_MSG("failed to allocate string array");
  return RCUTILS_RET_BAD_ALLOC;
}

When the allocation fails, the function returns RCUTILS_RET_BAD_ALLOC but leaves size set to the requested count while data stays NULL. The array then advertises entries that cannot be read. A caller that checks the return value is fine, but one that inspects the array afterwards — or passes it to code that iterates 0..size — walks a null pointer.

This PR sets size back to 0 on that error path, so a failed init leaves the array exactly as rcutils_get_zero_initialized_string_array() would.

Fixes # (no issue filed; found while reading the allocation failure paths)

Is this user-facing behavior change?

Only on the failure path. On success nothing changes. After a failed rcutils_string_array_init() the array is now consistently empty (data == NULL, size == 0) instead of reporting a non-zero size with no backing storage, and rcutils_string_array_fini() on it returns RCUTILS_RET_OK.

Did you use Generative AI?

Yes. Claude Opus 5, through Claude Code, found the inconsistent error path, wrote the one-line change in src/string_array.c, wrote the init_alloc_failure_leaves_array_empty test in test/test_string_array.cpp, and drafted this description. The build and test verification below was run afterwards and is reported exactly as the container printed it.

Additional Information

Verification. My earlier note said I had no ROS 2 workspace on this machine. That is now done, in the official ros:rolling-ros-base image with the test dependencies (osrf_testing_tools_cpp, performance_test_fixture) installed by rosdep. Environment: Ubuntu 26.04.1, ROS 2 Rolling, gcc 15.2.0, cmake 4.2.3. Base: a4cba34 on rolling.

With this change applied, colcon build --packages-select rcutils succeeds and the test file passes:

build : OK
test_string_array.gtest.xml : total=5 failures=0 errors=0
  init_alloc_failure_leaves_array_empty        passed

Then I reverted only src/string_array.c, keeping the new test, rebuilt, and it fails on current rolling:

build : OK
test_string_array.gtest.xml : total=5 failures=1 errors=0
  init_alloc_failure_leaves_array_empty        FAILED
    test_string_array.cpp:84
    Expected equality of these values:
      0u          Which is: 0
      sa.size     Which is: 3

The array reports three entries after an allocation that never happened. The other four tests in the file pass on both sides.