Re: autoload bug in 2.6-beta10

Re: autoload bug in 2.6-beta10

[P.Stephenson]

alex_ramos@wiltel.com wrote:
> 
> It appears that parameters are not passed to autoload functions
> in 2.6-beta10. Sorry if this had already been reported, but I
> joined the mailing list fairly recently.
> 
> e.g. 
> 	foo() { echo $1 }
> 
> If this function is autoloaded, then invoking "foo test" results
> in echoing a blank line.

Yes, I missed this when I wrote the ksh-compatible autoload code.  The
second problem---as noted by Zefram---is that that code seems to have
disappeared anyway (contrary to what the manual page still says).
Here's a new version of it actually for Richard's 2.6-test2
non-release.  It's a bit more reliable since I've explicitly tested
whether the defined function exists and has changed, which I did a
more roundabout way before.  I fixed the argument problem by having
doshfunc() not destroy the list, which requires a couple of extra
lines of code.

*** Src/exec.c.autoload	Fri Sep  8 17:18:07 1995
--- Src/exec.c	Mon Sep 11 14:31:43 1995
***************
*** 2286,2291 ****
--- 2286,2292 ----
  {
      List funcdef;
      char *nam;
+     int done = 0;
  
      if (errflag)
  	return;
***************
*** 2302,2315 ****
  	}
  	permalloc();
  	shf->flags &= ~PM_UNDEFINED;
! 	shf->funcdef = (List) dupstruct(funcdef);
  	heapalloc();
  	popheap();
      } else if (!shf->funcdef) {
  	return;
      }
  
!     doshfunc(shf->funcdef, cmd->args, shf->flags, 0);
  }
  
  /* execute a shell function */
--- 2303,2327 ----
  	}
  	permalloc();
  	shf->flags &= ~PM_UNDEFINED;
! 	funcdef = shf->funcdef = (List) dupstruct(funcdef);
  	heapalloc();
  	popheap();
+ 
+ 	doshfunc(shf->funcdef, cmd->args, shf->flags, 0);
+ 	/* See if this file defined the autoloaded function by name */
+ 	shf = (Shfunc) shfunctab->getnode(shfunctab, nam);
+ 	if (!shf || !shf->funcdef || shf->funcdef == funcdef) {
+ 	    /* It didn't:  that means we have already executed the
+ 	     * function itself.
+ 	     */
+ 	    done++;
+ 	}
      } else if (!shf->funcdef) {
  	return;
      }
  
!     if (!done)
! 	doshfunc(shf->funcdef, cmd->args, shf->flags, 0);
  }
  
  /* execute a shell function */
***************
*** 2347,2357 ****
  	opts[XTRACE] = OPT_SET;
      opts[PRINTEXITVALUE] = OPT_UNSET;
      if (doshargs) {
  	pparams = x =
  	    (char **) zcalloc(((sizeof *x) * (1 + countlinknodes(doshargs))));
! 	argzero = ztrdup(ugetnode(doshargs));
! 	while ((*x = (char *) ugetnode(doshargs)))
! 	    *x = ztrdup(*x), x++;
      } else {
  	pparams = (char **) zcalloc(sizeof *pparams);
  	argzero = ztrdup(argzero);
--- 2359,2371 ----
  	opts[XTRACE] = OPT_SET;
      opts[PRINTEXITVALUE] = OPT_UNSET;
      if (doshargs) {
+ 	LinkNode node = doshargs->first;
  	pparams = x =
  	    (char **) zcalloc(((sizeof *x) * (1 + countlinknodes(doshargs))));
! 	argzero = ztrdup((char *)node->dat);
! 	node = node->next;
! 	for (; node; node = node->next, x++)
! 	    *x = ztrdup((char *)node->dat);
      } else {
  	pparams = (char **) zcalloc(sizeof *pparams);
  	argzero = ztrdup(argzero);

-- 
Peter Stephenson <P.Stephenson@swansea.ac.uk>  Tel: +44 1792 205678 extn. 4461
WWW:  http://python.swan.ac.uk/~pypeters/      Fax: +44 1792 295324
Department of Physics, University of Wales, Swansea,
Singleton Park, Swansea, SA2 8PP, U.K.

Re: autoload bug in 2.6-beta10

[Zefram]

>It appears that parameters are not passed to autoload functions
>in 2.6-beta10. Sorry if this had already been reported, but I
>joined the mailing list fairly recently.
>
>e.g. 
>	foo() { echo $1 }
>
>If this function is autoloaded, then invoking "foo test" results
>in echoing a blank line.

This is not a bug.  At least, not entirely.  Autoloaded function files
should not contain the foo() bit, as the entire contents of the file is
taken as the contents of the function.  The first time you run the
undefined function, it just redefines itself.  The second time, it will
do what you want.  This is explained in section C5 of the Zsh FAQ.

But zsh is supposed to accept the Korn shell syntax, as of a few betas
ago, so there is a bug here.

-zefram