Update API tests to use native helper

[saxena-anurag]

Description

If two tests are using the same native module, there is a possible timing issue that the second test fails as the driver from first test case is not yet unloaded. There is already a test helper class native_module_helper_t that creates a copy of the native driver to enable tests to load it multiple times, but in some tests it is missed to use this helper, and the original driver is directly loaded.

This PR updates program_load_helper() helper in API tests to internally use native_module_helper_t so that each caller does not have to remember to do the same.

This issue was causing random failures in api_tests for ARM64 builds.

Testing

Existing CICD

Documentation

No

Installation

No

[shankarseal]

I would like the PR to take the following changes:

1. ebpf_native_load_driver must use ntstatus_to_ebpf_result from the status returned by ZwLoadDriver instead of returning a hard coded EBPF_FAILED and return using the EBPF_RETURN macro.
2. ebpf_native_load must log the error code returned by ebpf_native_load_driver.

[Copilot]

Pull request overview

This PR refactors API test helper functions to address timing issues when multiple tests use the same native module by moving native_module_helper_t logic into the program_load_helper() function. This eliminates the need for individual test cases to explicitly manage native module copies.

Key Changes:

• Integrated native_module_helper_t functionality directly into program_load_helper() with a new optional copy_file parameter
• Added helper functions _is_native_program() and _get_file_name_without_extension() to detect and process native driver files
• Simplified multiple test cases by removing explicit native_module_helper_t instantiations

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
tests/api_test/api_test.h	Added helper functions for native file detection and path processing; integrated native module copying logic into program_load_helper() with new copy_file parameter (defaults to true)
tests/api_test/api_test.cpp	Removed explicit native_module_helper_t usage from multiple test cases; updated tests to pass file names directly to helper functions; added copy_file=false parameter for special test cases that require loading the original file

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The function _is_native_program() has a bug when the filename doesn't contain a dot. If find_last_of(".") returns std::string::npos (which is typically SIZE_MAX), then substr(std::string::npos + 1) will wrap around to 0 and return the entire string, or could cause undefined behavior. You should check if find_last_of returns std::string::npos before calling substr.

Suggested change

	std::string file_extension = file_name_string.substr(file_name_string.find_last_of(".") + 1);
	size_t last_dot = file_name_string.find_last_of('.');
	std::string file_extension;
	if (last_dot != std::string::npos && last_dot + 1 < file_name_string.size()) {
	file_extension = file_name_string.substr(last_dot + 1);
	}

[Copilot]

When a native program file (.sys) is detected, the native_module_helper should be initialized with EBPF_EXECUTION_NATIVE instead of the passed execution_type. If execution_type is EBPF_EXECUTION_ANY, in non-JIT-disabled builds it won't resolve to EBPF_EXECUTION_NATIVE, causing the helper to incorrectly try to use a .o file instead of creating a copy of the .sys file. This breaks tests like test_ebpf_program_next_previous_native that pass .sys files but rely on program_load_helper being called with EBPF_EXECUTION_ANY. The initialization should use EBPF_EXECUTION_NATIVE when a .sys file is detected.

Suggested change

	_native_helper.initialize(file_name_without_extension.c_str(), execution_type);
	_native_helper.initialize(file_name_without_extension.c_str(), EBPF_EXECUTION_NATIVE);

[mikeagun]

I disagree with copilot - if invalid test parameters are used the test (or call within the test) should fail rather than being hidden within the helper. This comment looks like it points to a bug in the helper though.

[saxena-anurag]

I would like the PR to take the following changes:

1. ebpf_native_load_driver must use ntstatus_to_ebpf_result from the status returned by ZwLoadDriver instead of returning a hard coded EBPF_FAILED and return using the EBPF_RETURN macro.
2. ebpf_native_load must log the error code returned by ebpf_native_load_driver.

@shankarseal since this is a test only PR, is it fine to make the above 2 changes in a separate PR? I can file an issue to track the above 2 items.