Re: UserData.

[Edgar Toernig <froese@...>]

Hi,

Roberto Ierusalimschy wrote:
> 
> From: Edgar Toernig <froese@gmx.de>
> > Maybe this change
> >
> > -  ts->u.d.value = (udata == NULL) ? uts+1 : udata;
> > +  ts->u.d.value = s ? uts+1 : udata;
> >
> > would make a little bit more sense.  (There are other
> > problems too.  Still thinking about a solution.)
> 
> (The change is in lstring.c:122)
> 
> I think his fix will solve the problem, although I don't know what
> would be the "other problems".

For userdata?  Well, the LUA_ANYTAG stuff is IMHO bogus.  You get
some random objects back.  (I've simply removed it from Sol).  The
other thing is with lua_settag.  It's used to change the tag of
already existing userdata objects.  The only valid way to use this
is IMHO to invalidate a userdata object.  Any other usage may
generate duplicate objects with same udata/tag value and you no
longer know which one you get with lua_pushusertag.  So I made
an invalidate function (that's a little bit trivial at the moment)
and restricted the use of anonymous userdata objects (= userdata
with default tag) to valid pointers.

But the "other problems" were not only about userdata.  I was
twiddling with some other things at that time too.  Some user
function are pretty fragile.  Sometimes you may get unpredictable
results.  I.e. tinsert(x,foo()): if foo returns more than one value
you either get an error or a broken list.  Or dostring/dofile/call:
you don't know if the results come from the executed objects or from
the error handling.

Ciao, ET.