list-expand fix and a bug report

list-expand fix and a bug report

[Zefram]

-----BEGIN PGP SIGNED MESSAGE-----

The patch below makes list-expand refuse to do anything in
circumstances that currently cause trouble, such as inside braces.  In
fact, before this patch,

% echo {a,b<CTRL-G>

turns off my cursor, which is rather annoying.  (I have list-expand
bound to ^G.)

On an unrelated matter, I have found a problem with completion of filenames containing spaces.  If I type:

% touch 'spoon '
% echo spoon\ <TAB>

then in plain beta10 the space disappears after a couple of tabs, and
then the backslash goes, and it's a bit of a mess.  In an up-to-date
patched beta10, a space gets inserted *after* the cursor, and repeated
tabs insert more spaces without moving the cursor.  This happens in
beta10-hzoli10.1 too.  Can someone tell me which patch could have done
this, as my patch list doesn't show anything changing in this area.
More importantly, can anyone actually fix it?

Probably the same bug:

% mkdir t
% cd t
% touch spoon-
% touch 'spoon '
% echo spoon<TAB>

during menu completion, I get the sequence (_ representing cursor):

% echo spoon\_
% echo spoon-_

If I stop menu completion on the first of these, I find that there is a
space under the cursor.  This happens with plain beta10, patched beta10
and with beta10-hzoli10.1.  Also note that it only happens with space;
tab and other special characters are fine.

 -zefram

      *** Src/zle_tricky.c.old	Sat Jul 15 23:56:49 1995
      --- Src/zle_tricky.c	Sun Jul 16 02:07:47 1995
      ***************
      *** 1250,1260 ****
        	    feep();
        	goto end;
            }
      !     if (lst == COMP_LIST_EXPAND) {
      ! 	/* Only the list of exapansions was requested. */
      ! 	listlist(vl);
      ! 	goto end;
      !     } else if (peekfirst(vl) == (void *) s ||
        	       (!nextnode(firstnode(vl)) && *s == Tilde &&
        		(ss = dupstring(s), filesubstr(&ss, 0)) &&
        		!strcmp(ss, (char *)peekfirst(vl)))) {
      --- 1250,1256 ----
        	    feep();
        	goto end;
            }
      !     if (peekfirst(vl) == (void *) s ||
        	       (!nextnode(firstnode(vl)) && *s == Tilde &&
        		(ss = dupstring(s), filesubstr(&ss, 0)) &&
        		!strcmp(ss, (char *)peekfirst(vl)))) {
      ***************
      *** 1264,1269 ****
      --- 1260,1270 ----
        	    docompletion(s, COMP_COMPLETE, explincmd);
        	else
        	    feep();
      + 	goto end;
      +     }
      +     if (lst == COMP_LIST_EXPAND) {
      + 	/* Only the list of expansions was requested. */
      + 	listlist(vl);
        	goto end;
            }
            /* Remove the current word and put the expansions there. */

-----BEGIN PGP SIGNATURE-----
Version: 2.6.i

iQBVAgUBMAhzdWWJ8JfKi+e9AQGaJQH+Pg7cQ1Qez68CP6E8rl/GIsOcuwrmFiYy
KErcUDOujeMQuCetCaBieyobqAd7MbiHSuvErznEL4HmjT6KBLI/wQ==
=4YAu
-----END PGP SIGNATURE-----

Re: list-expand fix and a bug report

[Zoltan Hidvegi]

> ...
> On an unrelated matter, I have found a problem with completion of filenames containing spaces.  If I type:
> 
> % touch 'spoon '
> % echo spoon\ <TAB>
> 
> then in plain beta10 the space disappears after a couple of tabs, and
> then the backslash goes, and it's a bit of a mess.  In an up-to-date
> patched beta10, a space gets inserted *after* the cursor, and repeated
> tabs insert more spaces without moving the cursor.  This happens in
> beta10-hzoli10.1 too.  Can someone tell me which patch could have done
> this, as my patch list doesn't show anything changing in this area.
> More importantly, can anyone actually fix it?
> 
> Probably the same bug:
> 
> % mkdir t
> % cd t
> % touch spoon-
> % touch 'spoon '
> % echo spoon<TAB>
> 
> during menu completion, I get the sequence (_ representing cursor):
> 
> % echo spoon\_
> % echo spoon-_
> 
> If I stop menu completion on the first of these, I find that there is a
> space under the cursor.  This happens with plain beta10, patched beta10
> and with beta10-hzoli10.1.  Also note that it only happens with space;
> tab and other special characters are fine.

The patch below fixes all of these bugs and even some more:

% echo \f$V<TAB>  rsulted:
% echo \f$VENDO R

With the cursor under the `R'.

The patch is for beta10-hzoli10+. It can be applied to stock beta10. In that
case the rejection to hist.c should be ignored, and the rejection to
zle_tricky.c should be merged by hand in addx(). I'll make a
patch to stock beta10 if I'll have time.

Bye,
  Zoltan


--- 1.6	1995/07/17 22:00:18
+++ Src/hist.c	1995/07/18 22:37:00
@@ -116,7 +116,7 @@
 	 * history line where \ was used to escape the bang). So if     *
 	 * c == '\\' we fetch one more character to see if it's a bang, *
 	 * and if it is not, we unget it and reset c back to '\\'       */
-	inungetc(c), c = '\\';
+	safeinungetc(c), c = '\\';
     if (stophist || alstackind)
 	/* If the result is a bangchar which came from history or alias  *
 	 * expansion, we treat it as an escaped bangchar, unless history *
@@ -134,6 +134,15 @@
     return c;
 }
 
+/**/
+void
+safeinungetc(int c)
+{
+    if (lexstop)
+	lexstop = 0;
+    else
+	inungetc(c);
+}
 
 /* extract :s/foo/bar/ delimiters and arguments */
 
@@ -192,7 +201,7 @@
 	    int g = ingetc();
 
 	    if (g != bangchar)
-		inungetc(g);
+		safeinungetc(g);
 	    else {
 		qbang = 1;
 		return bangchar;
@@ -210,10 +219,7 @@
 	    return ingetc();
 	}
 	if ((!cflag && inblank(c)) || c == '=' || c == '(' || lexstop) {
-	    if (lexstop)
-		lexstop = 0;
-	    else
-		inungetc(c);
+	    safeinungetc(c);
 	    return bangchar;
 	}
 	cflag = 0;
--- 1.28	1995/07/17 20:36:06
+++ Src/zle_tricky.c	1995/07/18 23:00:44
@@ -54,9 +54,11 @@
 
 extern char *glob_pre, *glob_suf;
 
-/* we and wb hold the end/beginning position of the word we are completing */
+/* we and wb hold the end/beginning position of the word we are completing *
+ * offs is the cursor position within the tokenized current word after     *
+ * removing nulargs.                                                       */
 
-static int we, wb, usemenu, useglob;
+static int we, wb, offs, usemenu, useglob;
 
 /* A pointer to the current position in the menu-completion array (the one
    that was put in the command line last). */
@@ -550,7 +552,7 @@
     noerrs = ne;
     /* When completing between words, the contents of wb and we may be
        garbled. */
-    if (we > wb && inblank(line[we - 1]))
+    if (we > wb && inblank(line[we - 1]) && (we < 2 || line[we - 2] != '\\'))
 	we--;
     /* For vi mode, reset the start-of-insertion pointer to the beginning of
     the word being completed, if it is currently later.  Vi itself would never
@@ -772,7 +774,9 @@
 void
 addx(char **ptmp)
 {
-    if (!line[cs] || inblank(line[cs]) || line[cs] == ')' || line[cs] == '`') {
+    if (!line[cs] || line[cs] == '\n' ||
+	(iblank(line[cs]) && (!cs || line[cs-1] != '\\')) ||
+	line[cs] == ')' || line[cs] == '`') {
 	*ptmp = (char *)line;
 	line = (unsigned char *)halloc(strlen((char *)line) + 3);
 	memcpy(line, *ptmp, cs);
@@ -916,7 +920,8 @@
 	clwords[i] = ztrdup(tokstr);
 	sl = strlen(tokstr);
 	/* sometimes the lexer gives us token strings ending with spaces. */
-	while (sl && clwords[i][sl - 1] == ' ')
+	while (sl && clwords[i][sl - 1] == ' ' &&
+	       (sl < 2 || clwords[i][sl - 2] != Bnull))
 	    clwords[i][--sl] = '\0';
 	/* If this is the word the cursor is in and we added a `x',
 	   remove it. */
@@ -1083,12 +1088,17 @@
     /* This variable will hold the current word in quoted form. */
     qword = ztrdup(s);
     /* While buildin the quoted form, we also clean up the command line. */
+    offs = cs - wb;
     for (p = s, tt = qword, i = wb; *p; p++, tt++, i++)
 	if (INULL(*p)) {
+	    if (i < cs)
+		offs--;
 	    if (p[1] || *p != Bnull) {
-		if (*p == Bnull)
+		if (*p == Bnull) {
 		    *tt = '\\';
-		else {
+		    if (cs == i + 1)
+			cs++, offs++;
+		} else {
 		    ocs = cs;
 		    cs = i;
 		    foredel(1);
@@ -2162,7 +2172,7 @@
 {
     Compctl cc = NULL;
     int owe = we, owb = wb, ocs = cs, isf = 1;
-    int offs, t, sf1, sf2, ooffs;
+    int t, sf1, sf2, ooffs;
     char *p, *sd = NULL, sav, *tt, *s1, *s2, *os = NULL;
     unsigned char *ol = NULL;
 
@@ -2178,11 +2188,7 @@
 
     /* Go to the end of the word if complete_in_word is not set. */
     if (unset(COMPLETEINWORD) && cs != we)
-	cs = we;
-    /* offs holds the distance between the cursor and the beginning of the
-       word. */
-    if ((offs = cs - wb) > (t = strlen(s)))
-	offs = t;
+	cs = we, offs = strlen(s);
 
     ispattern = haswhat = lastambig = 0;
     patcomp = filecomp = NULL;
@@ -2233,6 +2239,7 @@
 	/* And adjust the word beginning/end variables. */
 	wb += sl;
 	we += sl - pl;
+	offs -= pl;
     }
     /* Does this compctl have a suffix (compctl -S)? */
     if ((ccsuffix = cc->suffix) && *ccsuffix) {
@@ -2258,12 +2265,18 @@
     }
     /* Do we have one of the special characters `~' and `=' at the
        beginning? */
-    if ((ic = *s) != Tilde && ic != Equals)
-	ic = (*s == '~' ? Tilde : (*s == '=' ? Equals : '\0'));
-    /* Re-adjust the offs variable after all the changes we might have done
-       to s, cs, and wb. */
-    if ((offs = cs - wb) > (t = strlen(s)))
-	offs = t;
+    switch (*s) {
+    case '~':
+    case Tilde:
+	ic = Tilde;
+	break;
+    case '=':
+    case Equals:
+	ic = Equals;
+	break;
+    default:
+	ic = '\0';
+    }
 
     /* Check if we have to complete a parameter name... */
 
@@ -2328,8 +2341,8 @@
 	    qparprelen = strlen(qparampre);
 	    *b = sav;
 	    /* And adjust wb, we, and offs again. */
-	    wb += qparprelen;
 	    offs -= b - s;
+	    wb = cs - offs;
 	    we = wb + e - b;
 	    s = b;
 	    /* And now make sure that we complete parameter names. */
@@ -3262,7 +3275,8 @@
 	}
 	if (menuend > ll)
 	    menuend = ll;
-	if (menuend && ((char)line[menuend - 1]) != singlec)
+	if (menuend && ((((char)line[menuend - 1]) != singlec) ||
+		(menuend > 1 && singlec == ' ' && line[menuend - 2] == '\\')))
 	    if (parampre && singlec == '/' && ((char)line[menuend]) == '/')
 		addedsuffix = 0;
 	    /* Now insert the slash or space if there is none already. */

Re: list-expand fix and a bug report

[Mark Borges]

>> Zoltan Hidvegi(ZH) wrote on Fri, 21 Jul 1995 20:29:30 +0100 (MET DST):
[...]
ZH> The patch below fixes all of these bugs and even some more:

Zoltan--

Do any of these patches for 10.1 (or 10+ for that matter) fix this:

------------------------------------------------------------------------------
>> Mark Borges(mb) wrote on Mon, 17 Jul 95 14:50:50 MDT:
mb> Are you seeing this?

mb> Somewhere between hzoli9 and hzoli10.1 <TAB> completion inside a for
mb> control structure doesn't work. For example,

mb> $ ~/bin/sparc/zsh-2.6h9 -f
mb> bjerknes% echo $ZSH_VERSION
mb> 2.6-beta9
mb> bjerknes% setopt autolist
mb> bjerknes% for i in a*; do
>> echo a<TAB>	# a tab correctly prints out the ambiguous files
mb> a1  a2  a3

mb> but in hzoli10.1:

mb> $ ~/bin/sparc/zsh-2.6h10.1 -f
mb> bjerknes% echo $ZSH_VERSION
mb> 2.6-beta10-hzoli10.1
mb> bjerknes% setopt autolist 
mb> bjerknes% for i in a*; do

>> echo a<TAB>	# Huh? Another a keeps getting added, and no list is printed.

>> echo aa<TAB>
      
>> echo aaa<TAB>	# can go on indefinitely
      
>> echo aaaa
      
>> echo aaaaa
      
>> echo aaaaaa
      
mb> This also has patches from zefram (I forget what they do). I was just
mb> curious if the bug has infected your copy as well.
------------------------------------------------------------------------------
Or, can anyone verify whether vanilla beta 10 has the above bug?

Thanks,

  -mb-

Re: list-expand fix and a bug report

[Zoltan Hidvegi]

Mark Borges wrote:
> Do any of these patches for 10.1 (or 10+ for that matter) fix this:

No, bat there is a patch below for that.

> >> Mark Borges(mb) wrote on Mon, 17 Jul 95 14:50:50 MDT:
> mb> Are you seeing this?
> 
> mb> Somewhere between hzoli9 and hzoli10.1 <TAB> completion inside a for
> mb> control structure doesn't work. For example,
> 
> mb> $ ~/bin/sparc/zsh-2.6h9 -f
> mb> bjerknes% echo $ZSH_VERSION
> mb> 2.6-beta9
> mb> bjerknes% setopt autolist
> mb> bjerknes% for i in a*; do
> >> echo a<TAB>	# a tab correctly prints out the ambiguous files
> mb> a1  a2  a3
> 
> mb> but in hzoli10.1:
> 
> mb> $ ~/bin/sparc/zsh-2.6h10.1 -f
> mb> bjerknes% echo $ZSH_VERSION
> mb> 2.6-beta10-hzoli10.1
> mb> bjerknes% setopt autolist 
> mb> bjerknes% for i in a*; do
> 
> >> echo a<TAB>	# Huh? Another a keeps getting added, and no list is printed.
> 
> >> echo aa<TAB>
>       
> >> echo aaa<TAB>	# can go on indefinitely
>       
> >> echo aaaa
>       
> >> echo aaaaa
>       
> >> echo aaaaaa
>       
> mb> This also has patches from zefram (I forget what they do). I was just
> mb> curious if the bug has infected your copy as well.

In fact the bug appeared after Peter's input patches. The problem again is
that unlike the old history code, ingetch() doesn't remove tokens from the
input stream.  I is not necessarily a bug although I think this should be put
back in there (it's only one line) as tokens in iput can cause completely
unexpected results.  But for now it is good as it showed up two old bug so
far.  This particular bug is in zle_tricky.c.  If the current line is part of
a command structure, zle prepends chline to the current line before calling
the lexer.  But chline contains HISTSPACEs which should be replaced with
normal spaces.  Here is the one line patch for that.  It can be (and should
be) applied to any zsh release.

Bye,

  Zoltan

--- 1.30	1995/07/22 22:52:16
+++ Src/zle_tricky.c	1995/07/22 23:00:34
@@ -522,7 +522,7 @@
 		*q++ = '\\';
 		*q++ = '\\';
 	    }
-	    *q++ = *p;
+	    *q++ = *p == HISTSPACE ? ' ' : *p;
 	}
 	/* Put a zero byte after it and append the original line contents. */
 	*q = '\0';