Fwd: [Pkg-zsh-devel] Bug#733075: zsh: command completion not working after semicolon

Fwd: [Pkg-zsh-devel] Bug#733075: zsh: command completion not working after semicolon

[Richard Hartmann]

Dear all,

bug confirmed:


% zsh -f
eudyptes% autoload  compinit;compinit
eudyptes% ls
bug  confirmed
eudyptes% echo zsh 5.0.4; <tab>
bug        confirmed


5.0.2 did not show this behaviour.


Richard

---------- Forwarded message ----------
From: Jay Berkenbilt <qjb@debian.org>
Date: Wed, Dec 25, 2013 at 3:42 AM
Subject: [Pkg-zsh-devel] Bug#733075: zsh: command completion not
working after semicolon
To: Debian Bug Tracking System <submit@bugs.debian.org>


Package: zsh
Version: 5.0.4-1
Severity: important

Usually when I type a command followed by ;, after the ;, zsh is back
to a state where TAB should complete on commands again.  This is
pretty basic completion behavior.  After I upgraded to 5.0.4-1, I
observed that this was no longer working.  As soon as I type a ;,
completion reverts to simple path completion.  I observe this with my
elaborate customizations and also with zsh -f.  I have not looked into
it, but it seems so fundamental, I thought it was worth reporting
right away in case there's a real issue here.

-- System Information:
Debian Release: jessie/sid
  APT prefers unstable
  APT policy: (500, 'unstable'), (200, 'experimental')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 3.11-2-amd64 (SMP w/8 CPU cores)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash

Versions of packages zsh depends on:
ii  libc6       2.17-97
ii  libcap2     1:2.22-1.2
ii  libtinfo5   5.9+20130608-1
ii  zsh-common  5.0.4-1

Versions of packages zsh recommends:
ii  libncursesw5  5.9+20130608-1
ii  libpcre3      1:8.31-2

Versions of packages zsh suggests:
ii  zsh-doc  5.0.4-1

-- no debconf information

_______________________________________________
Pkg-zsh-devel mailing list
Pkg-zsh-devel@lists.alioth.debian.org
http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-zsh-devel


-- 
Richard

Re: [Pkg-zsh-devel] Bug#733075: zsh: command completion not working after semicolon

[Richard Hartmann]

Dear all,

funnily enough, two semicolons fix this:

eudyptes% echo hi; echo hi; <tab>
zsh: do you wish to see all 4004 possibilities (1335 lines)?


Richard

Re: Bug#733075: Bug#733075: zsh: command completion not working after semicolon

[Frank Terbeck]

Hey Richard!

Richard Hartmann wrote:
> Seems we worked in parallel here; I already confirmed and forwarded to
> zsh-workers.

Seems like it. :-)

> 5.0.2-6 was not affected.

I git-bisected this to:

[568e0db7a964feefa45061967d0c7079a0e59c1e]
  31611: attempt to fix crash completing redirection in do loop

The relevant zsh-workers thread is:
  http://www.zsh.org/mla/workers/2013/msg00652.html


This is in zle_tricky.c, so naturally I'm afraid to touch it. :)

Regards, Frank

Re: Bug#733075: Bug#733075: zsh: command completion not working after semicolon

[Bart Schaefer]

On Dec 25, 12:01pm, Frank Terbeck wrote:
}
} I git-bisected this to:
} 
} [568e0db7a964feefa45061967d0c7079a0e59c1e]
}   31611: attempt to fix crash completing redirection in do loop
} 
} This is in zle_tricky.c, so naturally I'm afraid to touch it. :)

I think this is the right thing, but it would be helpful if someone
can also check that the redirection issue has not regressed.  (There
are no tests of command completion in Y01, so I'm not sure how to
add a regression test for these.)

diff --git a/Src/Zle/zle_tricky.c b/Src/Zle/zle_tricky.c
index 25f09c4..9d163ad 100644
--- a/Src/Zle/zle_tricky.c
+++ b/Src/Zle/zle_tricky.c
@@ -1071,7 +1071,7 @@ has_real_token(const char *s)
 static char *
 get_comp_string(void)
 {
-    enum lextok t0, tt0;
+    enum lextok t0, tt0, cmdtok;
     int i, j, k, cp, rd, sl, ocs, ins, oins, ia, parct, varq = 0;
     int ona = noaliases;
     /*
@@ -1146,6 +1146,7 @@ get_comp_string(void)
     linredir = inredir;
     zsfree(cmdstr);
     cmdstr = NULL;
+    cmdtok = NULLTOK;
     zsfree(varname);
     varname = NULL;
     insubscr = 0;
@@ -1264,6 +1265,7 @@ get_comp_string(void)
 	    ins = (tok == REPEAT ? 2 : (tok != STRING));
 	    zsfree(cmdstr);
 	    cmdstr = ztrdup(tokstr);
+	    cmdtok = tok;
 	    /* If everything before is a redirection, don't reset the index */
 	    if (wordpos != redirpos)
 		wordpos = redirpos = 0;
@@ -1271,10 +1273,11 @@ get_comp_string(void)
 	    /*
 	     * A following DOLOOP should cause us to reset to the start
 	     * of the command line.  For some reason we only recognise
-	     * DOLOOP for this purpose (above) if ins is set.  Why?
-	     * Don't ask pointless questions.
+	     * DOLOOP for this purpose (above) if ins is set.  Why?  To
+	     * handle completing multiple SEPER-ated command positions on
+	     * the same command line, e.g., pipelines.
 	     */
-	    ins = 1;
+	    ins = (cmdtok != STRING);
 	}
 	if (!lexflags && tt0 == NULLTOK) {
 	    /* This is done when the lexer reached the word the cursor is on. */

Re: Bug#733075: Bug#733075: zsh: command completion not working after semicolon

[Peter Stephenson]

On Wed, 25 Dec 2013 10:13:36 -0800
Bart Schaefer <schaefer@brasslantern.com> wrote:
> I think this is the right thing, but it would be helpful if someone
> can also check that the redirection issue has not regressed.  (There
> are no tests of command completion in Y01, so I'm not sure how to
> add a regression test for these.)

That was a crash when completing after the "<" in

  for f in 1; do < x

which was very easy to reproduce, and that's still working after the
change.

Regression tests for interactive features is another place needing a
volunteer (who wouldn't need to know about internals).  Felix did some
good work on Zle testing but it really needs a permanent maintainer.

get_comp_string() is screaming out for something less ad hoc, but I
don't know what that is apart from rewriting the shell to provide a
partial parse tree for a given context, which isn't going to happen.

pws

Re: Bug#733075: Bug#733075: zsh: command completion not working after semicolon

[Bart Schaefer]

On Jan 2,  6:16pm, Peter Stephenson wrote:
}
} That was a crash when completing after the "<" in
} 
}   for f in 1; do < x
} 
} which was very easy to reproduce, and that's still working after the
} change.

Right, and in fact I managed to work out a regression test for this
one as well (see the last two tests in Y01).

Re: Bug#733075: Bug#733075: zsh: command completion not working after semicolon

[Oliver Kiddle]

Peter Stephenson wrote:
> Regression tests for interactive features is another place needing a
> volunteer (who wouldn't need to know about internals).  Felix did some
> good work on Zle testing but it really needs a permanent maintainer.

It perhaps doesn't help that Felix's good work never got merged (I'm
assuming you're referring to workers/30103). I had missed that and took a
slightly different approach when adding a test for a vi-mode interactive
feature. Both approaches are useful: Felix's where the mini-buffer is
used and the other where buffer contents are more important. However, it
might be better if there isn't a clash in the names of the helper
functions so they could both be used in the same test script.

Oliver