From: Andreas K. <and...@ac...> - 2006-10-18 22:44:40
|
> > Jeff Hobbs wrote: > >> Also, it might be that we want to know both an absolute name for the > >> proc as well as the name used to call it. Perhaps another entry in > >> the dictionary, "called-as" ? > > > > > > In considering this, I realized that "called as" is actually the more > > important variant, > > Maybe. In most cases the "called as" value is available by applying > [lindex ... 0] to the "cmd" value. In most cases where it's not already > available that way, it will be fully qualified already (object names). > > because you can always do the absolute-izing by > > upleveling into that scope and making the necessary call. However, it > > would be very wrong to always provide a fully absolute name, as that may > > be more confusing than not. > > Judgment call, I guess. This stack trace data says there's a problem > when calling command "foo" at line X of file Y, and then you look at > line X of file Y and see it's the implementation of ::otherNs::original. > > Certainly potential for confusion, but I'd think that more data rather > than less would help clarify matters. A patch for the current implementation is up at http://sourceforge.net/tracker/index.php?func=detail&aid=1571568&group_id=10894& atid=310894 Done against Tcl 8.4 head, requires #define TCL_TIP280 to be active. Crash free on the extended testsuite (Linux/Intel). Should now be leak free too, managed to tack down my refcounting problem which caused compile-12.1 to fail. Experiment with this, and report your troubles, or where the output does not match expectations. I.e. it is now possible to try real and synthetic examples and discuss things grounded in reality. -- Andreas Kupries <andreask@ActiveState.com> Developer @ http://www.ActiveState.com Tel: +1 778-786-1122 |