[CEDET-devel] Re: no-next-method in SemanticDB

[drkm <dar...@ya...>]

"Eric M. Ludlam" writes:

> Was your other problem ever resolved that caused the need for this
> patch?  When EIEIO recompiles, it re-creates a new default superclass.
> Re-creation of any class is supposed to find objects that inherit from it
> and swap in the new value.  A great idea in principle, but sometimes
> it goes awry in the same way you had described the original problem.

  Not already.  I didn't already have time to try.  Maybe on
Thursday.  But the strange thing is that I already had the same
problem.  I'm on Cygwin here, and I had the same problem on a
Linux (at SF.net or lxplus.cern.ch, I don't remember), with the
pre2, I think.  Here with a CVS GNU Emacs, and there with a 21.3,
I think.  After installed CEDET the way the INSTALL file told me.

  I'll let you know if I find something interesting.

> An earlier question was how to distinguish which version of a method
> was the problem.  Each method is stored on a symbol the same name as
> the class.  Thus if method foo on object BAR  is called, you would see
> a stack that starts with foo, and eventually calls BAR, except BAR is
> not in the default symbol obarray.  As such, you can learn a lot from
> a stack... which is what was missing in this case.

  Yes.  Such a stack would be easily implementable.
'call-next-method' may easily maintains it, isn't it?

  One particularity of EIEIO (in regard of CLOS) is that the
dispatching is done only on the first argument.  So a stack of
classes symbols may be enough (the entire arguments lists, as I
first said, is too much).

  But having real objects representing methods could be of
interest, too.  They can be stored in the plist of the
same-name-of-the-class-symbols in the obarrays in the
'eieio-method-obarray' property of the generic function symbol,
for example.

> It should be possible to derive the active method symbol, and append
> the text together for the method name to make something
> comprehensible.

  What do you mean by "to derive the active method symbol"?

  Thanks,

--drkm