Re: [PATCH 2/3] terminal: set VMIN and VTIME in non-canonical mode

From: Junio C Hamano <gitster@pobox.com>
Date: Feb 16, 2022, 21:40 UTC
Message-ID: <xmqqpmnmcwwu.fsf@gitster.g>
In-Reply-To: <02009172e081ab5c4de8e2476c1142f97b41698e.1645008873.git.gitgitgadget@gmail.com>

"Phillip Wood via GitGitGadget" <gitgitgadget@gmail.com> writes:

Show 7 quoted lines

> From: Phillip Wood <phillip.wood@dunelm.org.uk>
>
> If VMIN and VTIME are both set to zero then the terminal performs
> non-blocking reads which means that read_key_without_echo() returns
> EOF if there is no key press pending. This results in the user being
> unable to select anything when running "git add -p".  Fix this by
> explicitly setting VMIN and VTIME when enabling non-canonical mode.

Makes sense.

Show 18 quoted lines

>
> Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
> ---
>  compat/terminal.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/compat/terminal.c b/compat/terminal.c
> index 20803badd03..d882dfa06e3 100644
> --- a/compat/terminal.c
> +++ b/compat/terminal.c
> @@ -57,6 +57,10 @@ static int disable_bits(tcflag_t bits)
>  	t = old_term;
>  
>  	t.c_lflag &= ~bits;
> +	if (bits & ICANON) {
> +		t.c_cc[VMIN] = 1;
> +		t.c_cc[VTIME] = 0;
> +	}

Quite sensible. I wonder if we can simplify the "are we looking at an ESC that is the first byte of a multi-byte control sequence?" logic in the caller by using inter-character delay, but it is probably better to go one step at a time like this patch does.

Show 13 quoted lines

>  	if (!tcsetattr(term_fd, TCSAFLUSH, &t))
>  		return 0;
>  
> @@ -159,7 +163,11 @@ static int disable_bits(DWORD bits)
>  
>  		if (bits & ENABLE_LINE_INPUT) {
>  			string_list_append(&stty_restore, "icanon");
> -			strvec_push(&cp.args, "-icanon");
> +			/*
> +			 * POSIX allows VMIN and VTIME to overlap with VEOF and
> +			 * VEOL - let's hope that is not the case on windows.
> +			 */
> +			strvec_pushl(&cp.args, "-icanon", "min", "1", "time", "0", NULL);

Interesting. So each call to read_key_without_echo() ends up being a run_command("stty -icanon min 1 time 0") followed by a read followed by another "stty".

At least while in -icanon mode, VEOF and VEOL do not take effect, and the potential overlap would not matter. It really depends on what happens upon restore.

Do we have similar "let's hope" on the tcsetattr() side, too?

>  		}
>  
>  		if (bits & ENABLE_ECHO_INPUT) {

Thanks.