skipping assignment and function statements; Interactive comments

skipping assignment and function statements; Interactive comments

[Rocky Bernstein]

I just added a skip command to zshdb (and committed to github), but it
seems that assignment statements and function definitions are not
really skipped.

Consider this example which is slightly modified from the regession
test that was put in when it was implimented:

  setopt localtraps localoptions debugbeforecmd
  trap '(( LINENO == 4 )) && setopt errexit' DEBUG
  x=three
  x=four  # Statement is supposed to be skipped
  print $LINENO five
  print x is $x

The output I get is:

  5 five
  x is four

And if the statement were a function definition, e.g.
  x() { return 4; }
I think x would get defined as a funciton

- - -
In tracking something else down I read this in zshoptions under
INTERACTIVE_COMMENTS:

   "Allow comments even in interactive shells".

I find this phrasing a little weird as it It seems to imply that one
never would want to write a comments, but sometimes one has to break
down and do it in an interactive shell. :-)

Perhaps the word "even" should be removed.

Re: skipping assignment and function statements; Interactive comments

[Peter Stephenson]

On Sat, 23 Aug 2008 00:15:31 -0400
"Rocky Bernstein" <rocky.bernstein@gmail.com> wrote:
> I just added a skip command to zshdb (and committed to github), but it
> seems that assignment statements and function definitions are not
> really skipped.

Yes, thanks, there's yet another infuriating special case of the sort
that makes me wish I'd picked an easier project.

That's the third change here; the stuff with intrap is because it
doesn't detect that it's already inside a trap (and hence should skip
debugging) until it gets into dotrapargs(), and in this case it seems
neater not to go to the trouble.  This will be even more sensible when I
get around to regenerating the line to be debugged at this point.

Index: Src/exec.c
===================================================================
RCS file: /cvsroot/zsh/zsh/Src/exec.c,v
retrieving revision 1.141
diff -u -r1.141 exec.c
--- Src/exec.c	22 Aug 2008 15:41:31 -0000	1.141
+++ Src/exec.c	25 Aug 2008 17:21:19 -0000
@@ -1068,7 +1068,7 @@
 		lineno = lnp1 - 1;
 	}
 
-	if (sigtrapped[SIGDEBUG] && isset(DEBUGBEFORECMD)) {
+	if (sigtrapped[SIGDEBUG] && isset(DEBUGBEFORECMD) && !intrap) {
 	    int oerrexit_opt = opts[ERREXIT];
 	    opts[ERREXIT] = 0;
 	    noerrexit = 1;
@@ -1086,11 +1086,12 @@
 	    donedebug = isset(ERREXIT) ? 2 : 1;
 	    opts[ERREXIT] = oerrexit_opt;
 	} else
-	    donedebug = 0;
+	    donedebug = intrap ? 1 : 0;
 
 	if (ltype & Z_SIMPLE) {
 	    next = state->pc + WC_LIST_SKIP(code);
-	    execsimple(state);
+	    if (donedebug != 2)
+		execsimple(state);
 	    state->pc = next;
 	    goto sublist_done;
 	}
Index: Test/C03traps.ztst
===================================================================
RCS file: /cvsroot/zsh/zsh/Test/C03traps.ztst,v
retrieving revision 1.14
diff -u -r1.14 C03traps.ztst
--- Test/C03traps.ztst	7 Aug 2008 16:25:16 -0000	1.14
+++ Test/C03traps.ztst	25 Aug 2008 17:21:19 -0000
@@ -415,6 +415,20 @@
 >3 three
 >5 five
 
+  # Assignments are a special case, since they use a simpler
+  # wordcode type, so we need to test skipping them separately.
+  fn() {
+    setopt localtraps localoptions debugbeforecmd
+    trap '(( LINENO == 4 )) && setopt errexit' DEBUG
+    x=three
+    x=four
+    print $LINENO $x
+    [[ -o errexit ]] && print "Hey, ERREXIT is set!"
+  }
+  fn
+1:Skip assignment from DEBUG trap
+>5 three
+
 %clean
 
   rm -f TRAPEXIT


-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/

Re: skipping assignment and function statements; Interactive comments

[Rocky Bernstein]

Thanks! Seems to work fine.

(Now I have to work out my own regression tests for on this ;-)

On Mon, Aug 25, 2008 at 1:22 PM, Peter Stephenson
<p.w.stephenson@ntlworld.com> wrote:
> On Sat, 23 Aug 2008 00:15:31 -0400
> "Rocky Bernstein" <rocky.bernstein@gmail.com> wrote:
>> I just added a skip command to zshdb (and committed to github), but it
>> seems that assignment statements and function definitions are not
>> really skipped.
>
> Yes, thanks, there's yet another infuriating special case of the sort
> that makes me wish I'd picked an easier project.
>
> That's the third change here; the stuff with intrap is because it
> doesn't detect that it's already inside a trap (and hence should skip
> debugging) until it gets into dotrapargs(), and in this case it seems
> neater not to go to the trouble.  This will be even more sensible when I
> get around to regenerating the line to be debugged at this point.
>
> Index: Src/exec.c
> ===================================================================
> RCS file: /cvsroot/zsh/zsh/Src/exec.c,v
> retrieving revision 1.141
> diff -u -r1.141 exec.c
> --- Src/exec.c  22 Aug 2008 15:41:31 -0000      1.141
> +++ Src/exec.c  25 Aug 2008 17:21:19 -0000
> @@ -1068,7 +1068,7 @@
>                lineno = lnp1 - 1;
>        }
>
> -       if (sigtrapped[SIGDEBUG] && isset(DEBUGBEFORECMD)) {
> +       if (sigtrapped[SIGDEBUG] && isset(DEBUGBEFORECMD) && !intrap) {
>            int oerrexit_opt = opts[ERREXIT];
>            opts[ERREXIT] = 0;
>            noerrexit = 1;
> @@ -1086,11 +1086,12 @@
>            donedebug = isset(ERREXIT) ? 2 : 1;
>            opts[ERREXIT] = oerrexit_opt;
>        } else
> -           donedebug = 0;
> +           donedebug = intrap ? 1 : 0;
>
>        if (ltype & Z_SIMPLE) {
>            next = state->pc + WC_LIST_SKIP(code);
> -           execsimple(state);
> +           if (donedebug != 2)
> +               execsimple(state);
>            state->pc = next;
>            goto sublist_done;
>        }
> Index: Test/C03traps.ztst
> ===================================================================
> RCS file: /cvsroot/zsh/zsh/Test/C03traps.ztst,v
> retrieving revision 1.14
> diff -u -r1.14 C03traps.ztst
> --- Test/C03traps.ztst  7 Aug 2008 16:25:16 -0000       1.14
> +++ Test/C03traps.ztst  25 Aug 2008 17:21:19 -0000
> @@ -415,6 +415,20 @@
>  >3 three
>  >5 five
>
> +  # Assignments are a special case, since they use a simpler
> +  # wordcode type, so we need to test skipping them separately.
> +  fn() {
> +    setopt localtraps localoptions debugbeforecmd
> +    trap '(( LINENO == 4 )) && setopt errexit' DEBUG
> +    x=three
> +    x=four
> +    print $LINENO $x
> +    [[ -o errexit ]] && print "Hey, ERREXIT is set!"
> +  }
> +  fn
> +1:Skip assignment from DEBUG trap
> +>5 three
> +
>  %clean
>
>   rm -f TRAPEXIT
>
>
> --
> Peter Stephenson <p.w.stephenson@ntlworld.com>
> Web page now at http://homepage.ntlworld.com/p.w.stephenson/
>