Re: [PATCH v2 1/3] t0301: fixes for windows compatibility

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 13, 2021, 18:09 UTC
Message-ID: <xmqqpmtcfiuq.fsf@gitster.g>
In-Reply-To: <nycvar.QRO.7.76.6.2109131343440.55@tvgsbejvaqbjf.bet>

Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:

Show 17 quoted lines

>> +test_path_is_socket () {
>> +	test -S "$1"
>> +}
>> +
>> +# in Windows, Unix Sockets look just like regular files
>> +uname_s=$(uname -s)
>> +case $uname_s in
>> +*MINGW*)
>> +	test_socket_exist=test_path_exists
>> +	;;
>> +*)
>> +	test_socket_exist=test_path_is_socket
>> +	;;
>> +esac
>
> A more canonical way would probably be to imitate what we do with `pwd` in
> `t/test-lib.sh`:

Thanks for bringing up a better practice.

Referring to a variable when calling a function gives a "we are doing something unusual" signal and it loses half its abstraction value at the callsites. E.g.

Show 5 quoted lines

>>  	test_when_finished "git credential-cache exit" &&
>> -	test -S "$XDG_CACHE_HOME/git/credential/socket" &&
>> +	$test_socket_exist "$XDG_CACHE_HOME/git/credential/socket" &&
>>  	test_path_is_missing "$HOME/.git-credential-cache/socket" &&
>>  	test_path_is_missing "$HOME/.cache/git/credential/socket"

I actually do not think it is so bad to just use test_path_exists without per-platform conditional in this case, but if we want to be more conservative, I agree with you that

	case ... in
	*MINGW*)
                test_path_is_socket () {
                        test_path_exists "$@"
                }
        	;;
	*)
                test_path_is_socket () {
                        test -S "$1"
                }
		;;
	esac

is the way to go.