[Bug translator/11556] New: @cast() with array indexing on pointers

["dsmith at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

While working on bug #9871, I tried to rewrite _get_fopv_size() from vfs.stp
from embedded-C into script language.

Here's the original embedded-C function:

function _get_fopv_size:long (iov:long, nr_segs:long)
%{ /* pure */
	struct iovec *iovp = (struct iovec *)(long)THIS->iov;
 	if (iovp) {
		int i;
		THIS->__retvalue = 0;
		for (i = 0 ; i < THIS->nr_segs ; i++)
			THIS->__retvalue += kread(&(iovp[i].iov_len));
	} else
		THIS->__retvalue = -1;

	CATCH_DEREF_FAULT();
%}

Here's my script language conversion:

function _get_fopv_size:long (iovp:long, nr_segs:long)
{
	if (iovp) {
		val = 0
		for (i = 0; i < nr_segs; i++)
			val += @cast(iovp, "iovec")[i]->iov_len
		return val
	}
	return -1
}

But this fails with an error:

semantic error: invalid access '[i]' for struct iovec: operator '[' at
/usr/local/share/systemtap/tapset/vfs.stp:134:31
        source: 			val += @cast(iovp, "iovec")[i]->iov_len
                			                           ^

I asked Josh, who implemented @cast(), about this, and he said:

> Right, well we actually do support array indexing on pointers.  The
> issue is a complication in the way @cast works.  When we normally do
> array access on pointers, we're looking at a type DIE of a pointer.  For
> the @cast, we're looking at the actual struct DIE rather than a pointer DIE.
>
> I think we could make this work though, perhaps adding a special-case
> for having an array access as the first component in @cast.  It
> shouldn't be too hard to do the pointer arithmetic first and then
> proceed normally. (Famous Last Words...)

So, I'd like @cast to be able to do array indexing on pointers.

-- 
           Summary: @cast() with array indexing on pointers
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: enhancement
          Priority: P2
         Component: translator
        AssignedTo: systemtap at sources dot redhat dot com
        ReportedBy: dsmith at redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=11556

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.