NetBSD Problem Report #60010

From www@netbsd.org  Mon Feb 16 09:56:17 2026
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
	(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
	 key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256
	 client-signature RSA-PSS (2048 bits) client-digest SHA256)
	(Client CN "mail.netbsd.org", Issuer "R13" (verified OK))
	by mollari.NetBSD.org (Postfix) with ESMTPS id 5B7021A9239
	for <gnats-bugs@gnats.NetBSD.org>; Mon, 16 Feb 2026 09:56:17 +0000 (UTC)
Message-Id: <20260216095616.3A6181A923C@mollari.NetBSD.org>
Date: Mon, 16 Feb 2026 09:56:16 +0000 (UTC)
From: thorpej@me.com
Reply-To: thorpej@me.com
To: gnats-bugs@NetBSD.org
Subject: "expandtab" behaves counter-intuitively (and different from vim)
X-Send-Pr-Version: www-1.0
X-From4GNATS: "thorpej@me.com via gnats" <gnats-admin@NetBSD.org>

>Number:         60010
>Category:       bin
>Synopsis:       "expandtab" behaves counter-intuitively (and different from vim)
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    bin-bug-people
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Mon Feb 16 10:00:01 +0000 2026
>Last-Modified:  Tue Feb 17 06:45:00 +0000 2026
>Originator:     Jason Thorpe
>Release:        NetBSD 10.99.10 (but 11 is the same)
>Organization:
Formatting Fanatics Anonymous
>Environment:
NetBSD the-ripe-vessel 10.99.10 NetBSD 10.99.10 (GENERIC) #68: Fri Nov 17 14:40:27 PST 2023  thorpej@the-ripe-vessel:/space/src/sys/arch/amd64/compile/GENERIC amd64
>Description:
"expandtab" is documented as:

       expandtab, et [off]
              Prevent the use of <tab> characters in leading whitespace when
              shifting text, autoindenting, indenting with <control-T>, or
              outdenting with <control-D>.

This is different behavior from vim, for which <control-I> (i.e. <tab>) also applies.  As you know, vim is shipped as the "vi" on a pretty large number of systems, thus making NetBSD's behavior in this regard an outlier.

I think the documented behavior is counter-intuitive and "correct" only in the sense that it matches the documentation.  It leads to inconsistent indenting within a source file because <tab> is an easy (and natural) character to type, whereas <control-T> is not.  Furthermore, the setting is called "expandtab" and the key on the keyboard labeled "TAB" does not get the "expand" treatment.
>How-To-Repeat:
Edit a file.  Set "ts=2", "et", and "ai".  Create a line with a leading <tab> character and some text.  Hit return and let the editor auto-indent the next line and type some additional text.  Now, set "ts=8" and observe.

  oink
  snort

        oink
  snort

>Fix:
Yes please!

>Audit-Trail:
From: Kimmo Suominen <kim@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and
 different from vim)
Date: Mon, 16 Feb 2026 12:54:46 +0200

 Hi,

 I don't think this is a bug. Indenting is done with Control-T, not
 Control-I (TAB).  Inputting a TAB is done with the TAB key or with
 Control-I.  It would be counter-intuitive if TAB would produce spaces
 instead.

 I think vim has gotten this wrong (among other things).

 However, it is easy to install vim from pkgsrc, and pkg_alternatives
 can be used to make it available as vi (as one possible approach).

 Kind regards,
 + Kimmo

From: Kimmo Suominen <kim@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and
 different from vim)
Date: Mon, 16 Feb 2026 13:03:05 +0200

 Or possibly

   map! ^I ^T

 in your ~/.exrc might solve it (but can also be annoying).

 Cheers,
 + Kimmo

From: Jason Thorpe <thorpej@me.com>
To: gnats-bugs@netbsd.org
Cc: gnats-admin@netbsd.org,
 netbsd-bugs@netbsd.org
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and different
 from vim)
Date: Mon, 16 Feb 2026 08:03:32 -0500

 > I don't think this is a bug. Indenting is done with Control-T, not
 > Control-I (TAB).  Inputting a TAB is done with the TAB key or with
 > Control-I.  It would be counter-intuitive if TAB would produce spaces
 > instead.
 >=20
 > I think vim has gotten this wrong (among other things).

 It just seems a ridiculous from a UI perspective.  Manual indenting is =
 done *a lot*, and Control-T is simply not a convenient key combination =
 (no matter where you keep your Control key) where as the Tab key is =
 *right there* and I=E2=80=99d venture to guess that practically everyone =
 instinctively hits Tab (unless you=E2=80=99re an emacs user, in which =
 case you=E2=80=99re a lost cause :-).  (The subtext here is: =E2=80=9CIf =
 I wanted to torture myself with bucky bits, I=E2=80=99d use emacs!=E2=80=9D=
 ).  Having consistency in formatting should not require setting both ts =
 and sw and overcoming decades of muscle memory to hit an awkward key =
 combination.

 -- thorpej

From: Valery Ushakov <uwe@stderr.spb.ru>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and
 different from vim)
Date: Mon, 16 Feb 2026 18:54:34 +0300

 On Mon, Feb 16, 2026 at 13:05:01 +0000, Jason Thorpe via gnats wrote:

 > Having consistency in formatting should not require setting both ts
 > and sw and overcoming decades of muscle memory to hit an awkward key
 > combination.

 If your muscle memory is that of vim, then use vim.  Consider that
 other people may have different muscle memory and you are proposing to
 ruin theirs.

 Full disclosure: I'm an emacs user.  Bite me.

 -uwe

From: Jason Thorpe <thorpej@me.com>
To: gnats-bugs@netbsd.org
Cc: gnats-admin@netbsd.org,
 netbsd-bugs@netbsd.org
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and different
 from vim)
Date: Mon, 16 Feb 2026 11:20:34 -0500

 > On Feb 16, 2026, at 10:55=E2=80=AFAM, Valery Ushakov via gnats =
 <gnats-admin@NetBSD.org> wrote:
 >=20
 > If your muscle memory is that of vim, then use vim.  Consider that
 > other people may have different muscle memory and you are proposing to
 > ruin theirs.

 Well, this really only comes into play when you=E2=80=99re working on a =
 code base that doesn=E2=80=99t conform to KNF (were the Tab vs Control-T =
 distinction is pretty much meaningless).  *This* is where my muscle =
 memory comes from, 30 years of writing/editing KNF-conforming code with =
 the default ts=3D8, where hitting the Tab key absolutely does the right =
 thing and Control-T simply isn=E2=80=99t necessary.

 Like, it=E2=80=99s been so =E2=80=9Cout of sight out of mind=E2=80=9D =
 that I had completely forgotten about the sw property because with =
 KNF-conforming code you simply don=E2=80=99t have to care.

 The other thing that irks me about the current situation: the setting in =
 question is called =E2=80=9Cexpandtab=E2=80=9D, not =E2=80=9Cexpandcontrol=
 t=E2=80=9D to =E2=80=9Cexpandtoshiftwidth=E2=80=9D.

 The issue only came up for me recently because I=E2=80=99ve been working =
 on a code base that is in a sort of hybrid GNU/KNF style, which uses =
 2-space tabs and all indents are expanded to spaces.  My assumption is =
 that the original author wrote it using emacs, although none of the =
 usual emacs formatting markup is present in any of the files.

 In any case, I would be happy enough with something I could put into my =
 ~/.exrc (which for the last 30 years has read "set ai nomagic sm=E2=80=9D)=
  to enable what I would term =E2=80=9Crational expandtab behavior=E2=80=9D=
 .  That way no one else=E2=80=99s muscle memory would be disturbed.

 -- thorpej

From: RVP <rvp@SDF.ORG>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: bin/60010: "expandtab" behaves counter-intuitively (and different
 from vim)
Date: Tue, 17 Feb 2026 06:40:20 +0000 (UTC)

 This OK? (`:set compat' is on so that I don't get yelled at...too much.)

 -RVP

 ---START patch---
 diff -urN external/bsd/nvi.orig/dist/common/options.c external/bsd/nvi/dist/common/options.c
 --- external/bsd/nvi.orig/dist/common/options.c	2019-10-04 09:01:59.000000000 +0000
 +++ external/bsd/nvi/dist/common/options.c	2026-02-17 05:22:52.312044929 +0000
 @@ -80,6 +80,8 @@
   	{L("combined"),	NULL,		OPT_0BOOL,	OPT_NOSET|OPT_WC},
   /* O_COMMENT	  4.4BSD */
   	{L("comment"),	NULL,		OPT_0BOOL,	0},
 +/* O_COMPAT	  NetBSD 12.0 */
 +	{L("compat"),	NULL,		OPT_1BOOL,	0},
   /* O_TMP_DIRECTORY	    4BSD */
   	{L("directory"),	NULL,		OPT_STR,	0},
   /* O_EDCOMPATIBLE   4BSD */
 @@ -282,6 +284,7 @@
   	{L("aw"),	O_AUTOWRITE},		/*     4BSD */
   	{L("bf"),	O_BEAUTIFY},		/*     4BSD */
   	{L("co"),	O_COLUMNS},		/*   4.4BSD */
 +	{L("cp"),	O_COMPAT},		/*   NetBSD 12.0 */
   	{L("dir"),	O_TMP_DIRECTORY},	/*     4BSD */
   	{L("eb"),	O_ERRORBELLS},		/*     4BSD */
   	{L("ed"),	O_EDCOMPATIBLE},	/*     4BSD */
 diff -urN external/bsd/nvi.orig/dist/docs/vi.man/vi.1 external/bsd/nvi/dist/docs/vi.man/vi.1
 --- external/bsd/nvi.orig/dist/docs/vi.man/vi.1	2025-04-08 04:00:22.996066228 +0000
 +++ external/bsd/nvi/dist/docs/vi.man/vi.1	2026-02-17 05:55:07.869829433 +0000
 @@ -1087,6 +1087,22 @@
   only.
   Skip leading comments in shell, C and C++ language files.
   .TP
 +.B "compat, cp [on]"
 +.I \&Vi
 +only.
 +Toggle compatibility with traditional
 +.I \&Vi
 +behaviour.
 +Only effects the behaviour of
 +.BR "autoindent"
 +and
 +.BR "expandtab",
 +where setting
 +.BR "nocompat"
 +will expand
 +.I all
 +tabs to spaces; ie. not only those at the beginning of lines.
 +.TP
   .B "directory, dir [environment variable TMPDIR, or /tmp]"
   The directory where temporary files are created.
   .TP
 diff -urN external/bsd/nvi.orig/dist/motif_l/m_options.c external/bsd/nvi/dist/motif_l/m_options.c
 --- external/bsd/nvi.orig/dist/motif_l/m_options.c	2014-01-26 21:43:45.000000000 +0000
 +++ external/bsd/nvi/dist/motif_l/m_options.c	2026-02-17 05:14:23.065699319 +0000
 @@ -100,6 +100,7 @@
   	{ optString,	"path",		},
   	{ optTerminator,		},
   }, general[] = {
 +	{ optToggle,	"compat",	},
   	{ optToggle,	"exrc",		},
   	{ optToggle,	"lisp",		},
   	{ optToggle,	"modeline",	},
 diff -urN external/bsd/nvi.orig/dist/vi/v_txt.c external/bsd/nvi/dist/vi/v_txt.c
 --- external/bsd/nvi.orig/dist/vi/v_txt.c	2020-07-07 10:58:43.000000000 +0000
 +++ external/bsd/nvi/dist/vi/v_txt.c	2026-02-17 05:25:13.816201508 +0000
 @@ -1201,6 +1201,10 @@
   		if (FL_ISSET(is_flags, IS_RUNNING))
   			FL_SET(is_flags, IS_RESTART);
   		break;
 +	case K_TAB:
 +		if (O_ISSET(sp, O_COMPAT))
 +			goto ins_ch;
 +		/* FALLTHROUGH */
   	case K_CNTRLT:			/* Add autoindent characters. */
   		if (!LF_ISSET(TXT_CNTRLT))
   			goto ins_ch;
 ---END patch---

Home	PR Database Search