Add support for Catch2 v3

[bluecube]

As far as I can tell, v3 binaries don't work with the current Catch2 support, because --list-test-names-only command line option is missing.

[nrbnlulu]

Does your tests gets collected?

[bluecube]

No, it doesn't.

[nrbnlulu]

just crafted this minimal thing

import json
import re
import subprocess
from functools import cached_property
from pathlib import Path
from typing import Optional

import pytest
from typing_extensions import TypedDict

build_dir = Path(__file__).parent.parent / "build"


class TestProperty(TypedDict):
    name: str
    value: str


class TestDefinition(TypedDict):
    name: str
    properties: list[TestProperty]


class CtestDiscoveryResult(TypedDict):
    tests: list[TestDefinition]


class CtestTestCommand:
    def __init__(self, test_name: str) -> None:
        self._data: list[str] = ["ctest", "-R"]
        self.test_name: str = test_name
        self.ret_res: Optional[subprocess.CompletedProcess] = None

    def add_command(self, command: str):
        self._data.append(command)

    @cached_property
    def failed_log(self) -> str | None:
        assert self.ret_res
        try:
            match = re.findall("(in: )(.*)(\\.log)", self.ret_res.stderr.decode())
            log_file = match[0][1]
        except IndexError:
            return None
        return Path(log_file + ".log").resolve(True).read_text()

    def run(self) -> None:
        self.ret_res = subprocess.run(
            [*self._data, self.test_name],
            cwd=build_dir.resolve(True),
            capture_output=True,
        )


def collect_tests() -> list[CtestTestCommand]:
    ret: list[CtestTestCommand] = []
    res = subprocess.run(
        ["ctest", "--show-only=json-v1"],
        cwd=build_dir.resolve(True),
        capture_output=True,
    )
    data: CtestDiscoveryResult = json.loads(res.stdout)
    for test in data["tests"]:
        ret.append(CtestTestCommand(test["name"]))
    return ret


test_commands = collect_tests()


@pytest.mark.parametrize("command", test_commands, ids=lambda v: v.test_name)
def test_cpp(command: CtestTestCommand):
    command.run()
    if log_file := command.failed_log:
        pytest.fail(msg=f"\n {'-'*8} Test {command.test_name} Failed {'-'*8} \n {log_file}")

Just point it to the build dir and it would do the rest.

[bluecube]

Wow you're fast :) . Thank you. I hope I'll be able to try it out later today.

[nicoddemus]

If it works, happy to review/merge a PR. 👍

[nrbnlulu]

@nicoddemus I don't use any library code though, I just find tests and run them via parameterize
something close to what I suggested in #20

[nicoddemus]

No worries, but I like to mention that to encourage someone interested in Catch2 v3 to contribute.

[bluecube]

@nrbnlulu Your code didn't work for me directly, because I don't use ctest in my project, I needed to call the test binary directly. But it gave me some starting point and I ended up modifying it for my use case ( https://github.com/bluecube/tablog/blob/87506b3b69a873914d6ea726facf4c6c4bf752ec/encoder/tests/test_adapter.py ), so thank you for the push in the right-ish direction :)

I'd still like to have more complete support for v3, though, including specific failed assertions and random generator seed, so I'll try to make a PR.

Should I add Catch2v3Facade, or add it to the existing one?

[nicoddemus]

Should I add Catch2v3Facade, or add it to the existing one?

Whichever is more convenient: if the changes are small and localized, might make sense to add to the existing facade, however if you need to change every single method to do something different, it might make more sense to write a new facade.

[nrbnlulu]

@bluecube AFAIK if you use catch with Cmake it generates ctest binary for you am I wrong?

[bluecube]

That might be true, but I'm using it with SCons.

[jnewb1]

closed by: #115