Re: [PATCH] help.c: help.autocorrect=prompt waits for user action

2 messages, 2 authors, 2021-08-14 · open the first message on its own page

Re: [PATCH] help.c: help.autocorrect=prompt waits for user action

From: Junio C Hamano <hidden>
Date: 2021-08-12 19:38:03

quoted hunk
@@ -618,7 +622,18 @@ const char *help_unknown_cmd(const char *cmd)
 				   _("Continuing under the assumption that "
 				     "you meant '%s'."),
 				   assumed);
-		else {
+		else if (autocorrect == AUTOCORRECT_PROMPT) {
+			if (!isatty(STDIN_FILENO) | !isatty(STDERR_FILENO))
Don't do bitwise or when you only care about boolean outcome.
+				exit(1);
This is not a new issue, but I think it is probably a bug in the
original design of the help-unknown that we do not do this check
much early in help_unknown_cmd(), to disable auto-correction in
a non-interactive session.  If we did so, we do not have to have
this check here. 

If we cannot yet come to consensus that disabling autocorrection
when running non-interactively is a good idea, at least we should be
able to do so only for _PROMPT case, like the attached patch at the
end.
+			char *answer;
This is decl-after-statement, but as I am suggesting to get rid of
the tty check done inside this block, it will become OK.
+			fprintf_ln(stderr, _("Assuming you meant: '%s'."),
+				   assumed);
+			answer = git_prompt(_("Continue? (y/N)"), PROMPT_ECHO);
Hmph, the above does not look WRONG per-se, but I wonder if it is
easier for the users to see a single line, e.g.

	struct strbuf msg = STRBUF_INIT;
	strbuf_addf(&msg, _("Run '%s' instead? (y/N)"), assumed);
        answer = git_prompt(msg.buf, PROMPT_ECHO);
        strbuf_release(&msg);

Thanks.

diff --git i/help.c w/help.c
index e22ba1d246..3629cb8796 100644
--- i/help.c
+++ w/help.c
@@ -540,6 +540,13 @@ const char *help_unknown_cmd(const char *cmd)
 
 	read_early_config(git_unknown_cmd_config, NULL);
 
+	/*
+	 * Disable autocorrection prompt in a non-interactive session
+	 */
+	if ((autocorrect != AUTOCORRECT_PROMPT) &&
+	    (!isatty(0) || !isatty(2)))
+		autocorrect = AUTOCORRECT_NEVER;
+
 	if (autocorrect == AUTOCORRECT_NEVER) {
 		fprintf_ln(stderr, _("git: '%s' is not a git command. See 'git --help'."), cmd);
 		exit(1);


Re: [PATCH] help.c: help.autocorrect=prompt waits for user action

From: Azeem Bande-Ali <hidden>
Date: 2021-08-14 03:07:57

On Thu, Aug 12, 2021 at 3:38 PM Junio C Hamano [off-list ref] wrote:
If we cannot yet come to consensus that disabling autocorrection
when running non-interactively is a good idea, at least we should be
able to do so only for _PROMPT case, like the attached patch at the
end.
Thanks for the suggestion. I think it makes sense to skip the autocorrection
completely when the setting is "prompt" and the user is not running
interactively.
quoted
+                     fprintf_ln(stderr, _("Assuming you meant: '%s'."),
+                                assumed);
+                     answer = git_prompt(_("Continue? (y/N)"), PROMPT_ECHO);
Hmph, the above does not look WRONG per-se, but I wonder if it is
easier for the users to see a single line, e.g.

        struct strbuf msg = STRBUF_INIT;
        strbuf_addf(&msg, _("Run '%s' instead? (y/N)"), assumed);
        answer = git_prompt(msg.buf, PROMPT_ECHO);
        strbuf_release(&msg);
I looked at the behavior for other settings and they
all behave pretty compactly. So I think this suggested
solution does fit better.
+       /*
+        * Disable autocorrection prompt in a non-interactive session
+        */
+       if ((autocorrect != AUTOCORRECT_PROMPT) &&
+           (!isatty(0) || !isatty(2)))
+               autocorrect = AUTOCORRECT_NEVER;
+
That should be `(autocorrect == AUTOCORRECT_PROMPT)` right?
So we skip the autocorrect if the user is not running interactively
and has the "prompt" setting.

Thanks for all the feedback! I will send out a patch incorporating
this feedback soon.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help