Small fixes

[drebelsky]

• Add tracy binaries to .gitignore
• Add check for SecretKey::random in tests
• Remove duplicated header guard
• Don't run Soroban tests when Catch2 tests fail
• Remove outdated documentation (Remove support for old V1 survey #4660 removed the "surveytopology" endpoint)

[Copilot]

Pull request overview

This PR includes several small miscellaneous fixes to improve the development workflow and test infrastructure.

Changes:

• Adds tracy profiling tool binaries to .gitignore
• Adds static check to prevent use of SecretKey::random() in test files
• Removes redundant #pragma once directive from basen.h header (which already has include guards)
• Attempts to modify CI to skip Soroban tests when Catch2 tests fail

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 1 comment.

File	Description
.gitignore	Adds tracy-capture and tracy-gui binaries to ignore list
src/test/check-nondet	Adds check to prevent non-deterministic SecretKey::random() usage in tests
lib/util/basen.h	Removes duplicate #pragma once (file uses traditional include guards)
ci-build.sh	Adds TESTS_ENVIRONMENT to skip tests after failures (has implementation issues)

[Copilot]

The TESTS_ENVIRONMENT variable in automake is designed to set environment variables that will be available to test scripts, not to execute conditional shell logic that gates test execution. The approach here of using [ $$failed -ne 0 ] && exit 77 || may not work as intended because:

1. The failed variable in automake's serial test harness is a shell variable in the test-suite.log generation logic, not an environment variable accessible to individual test commands
2. Even if failed were accessible, the TESTS_ENVIRONMENT is typically evaluated before the test command in a way that may not interact correctly with the test exit codes

Consider using a custom test wrapper script or LOG_COMPILER variable instead to implement this skip-on-failure logic. A wrapper script approach would be more reliable and easier to understand.

Suggested change

	# For automake serial tests, once one test fails, stop executing further tests;
	# exit code 77 indicates skipped tests
	export TESTS_ENVIRONMENT='[ $$failed -ne 0 ] && exit 77 || '
	# failed is not defined for parallel tests; to avoid problems with an unset
	# variable in TESTS_ENVIRONMENT, we define it here
	export failed=0

[drebelsky]

Putting some notes on the CI modification. This approach seems to work, although it does feel somewhat hacky (especially since it's relying on the internal automake implementation of serialized tests using a failed variable). In particular, I don't think it will handle stopping the rest of the tests if one of the recursive lib tests fails (although, if the lib tests are run serially, it will stop that serial run). Additionally, since the order of the tests in src is selftest-{no}pg, check-nondet, check-sorobans, we get the odd behavior where failing check-nondet stops check-sorobans from running (but not selftest). I'm happy to remove or rewrite the commit based on what others think, but I'm leaning toward wrapping selftest-{no}pg and check-sorobans into a wrapper script so that check-sorobans only runs when selftest passes (leaving the rest of the behavior as it was pre this PR).

[bboston7]

I'm leaning toward wrapping selftest-{no}pg and check-sorobans into a wrapper script so that check-sorobans only runs when selftest passes (leaving the rest of the behavior as it was pre this PR).

I like this option. I agree that depending on automake internals is not ideal.