Fixes #7350. Added support for absolute paths for "forge test --match-path"

[matthewliu10]

Motivation

Related to #7350 .

forge test --match-path absolute_path will report No tests match the provided pattern, only relative path works.

--match-path should work for both absolute path and relative path.

Solution

If the path given by the user is absolute, and the path given to the function is relative, it uses the canonicalize() function to turn the path given to the function into an absolute path so that the paths can be compared.

Notes

I initially tried converting both paths into their canonical forms using canonicalize and directly comparing them, however this failed on the can_test_with_match_path test because it would attempt to canonicalize "*src/ATest.t.sol" which causes an error.

There may be cases where the paths refer to the same location but is_match() still returns false. For example, if the user inputs "/home/user/a/b/../..//c/", it may not match with "/home/user/c/" even though they are equivalent. Note: canonicalize() resolves any symbolic links, relative path components, and removes any redundant components in the path.

Tests

The following 5 tests fail when running cargo test --all --all-features:

cmd::can_install_missing_deps_build cmd::can_install_missing_deps_test cmd::can_update_library_with_outdated_nested_dependency config::can_parse_default_fs_permissions ext_integration::snekmate

However, the following 4 tests still failed when I reverted the changes that I made (reverted code to the original code that I pulled from the repo):

cmd::can_install_missing_deps_build cmd::can_install_missing_deps_test cmd::can_update_library_with_outdated_nested_dependency ext_integration::snekmate

The following test failed when I ran cargo test --all --all-features, but passed when I ran it individually:

config::can_parse_default_fs_permissions

[matthewliu10]

The test that is failing is the new one that I added. When testing locally, it is passing:

[matthewliu10]

I've just run the failing tests locally. Every test either fails when using the code from the repo, or passes when using both the code from the repo and the new code:

forge::cli config::can_prioritise_closer_lib_remappings (fails with original code and new code) forge::cli create::can_create_using_unlocked (fails with original code) forge::cli config::can_detect_lib_foundry_toml (fails with original code) forge::cli config::can_override_config (fails with original code) forge::cli create::can_create_with_constructor_args (fails with original code) forge::cli cmd::can_init_vscode (fails with original code) anvil::it traces::test_trace_address_fork (passes with original code and new code) anvil::it fork::test_fork_call (passes with original code and new code) anvil::it fork::test_fork_timestamp (passes with original code and new code) anvil::it otterscan::can_call_ots_get_internal_operations_contract_selfdestruct (passes with original code and new code) anvil::it traces::test_trace_address_fork2 (passes with original code and new code) forge::cli create::can_create_template_contract (fails with original code) chisel::cache test_write_session (passes with original code and new code)

[klkvr]

yeah, the ci is a little bit broken right now after forge-std release, we will fix this soon

[mattsse]

is fixed on master

[matthewliu10]

I just synced it again

[matthewliu10]

I just ran the following tests locally before and after syncing with master. They all passed with both versions.

forge::cli script::can_broadcast_script_skipping_simulation forge::cli test_cmd::can_disable_block_gas_limit anvil::it traces::test_trace_address_fork anvil::it fork::test_fork_timestamp

[matthewliu10]

Please review when you get the chance

[zerosnacks]

Hi @matthewliu10

Thanks for your PR! I've pushed some changes to fix the merge conflict (glob.rs was moved) and some additional tests

[DaniPopes]

pending @klkvr