CEDET

>>> David <de_bb@...> seems to think that:
>I'm using the latest CEDET from CVS with Emacs 22.2. 
>
>Sometimes I get the following error:
>
>Idle Parse Error: "#<buffer XYZ> - Wrong type argument: (or object-p class-p), nil" [3 times]
>
>or this one:
>
>eieio-oref: Wrong type argument: (or object-p class-p), nil
>
>I'd love to give a full backtrace, but somehow debug-on-error doesn't do
>the trick here. 

I've been trying to find and track these down too.  You can use:

M-x semantic-debug-idle-function

or

M-x semantic-debug-idle-work-function

will produce a nice stack.  Thanks.

>Another problem is with smart completion in ELisp buffers. If I load an
>ELisp file, it gets correctly scanned and I can see all the variables,
>defuns, etc. in the TAGS menu. I can also navigate with the senator
>jumping commands, but the smart completion doesn't work until I do an
>M-x eval-buffer - then, suddenly, everything works as expected. Is this
>the wanted behaviour? Do I first have to eval the buffer to get smart
>completion to work?
  [ ... ]

The Emacs Lisp support has both the parsing part, and an "omniscient"
database that queries Emacs' known symbols.  It seems likely that the
include path technique isn't working quite right, but when you eval
the buffer, the omniscient database starts working, and lets symbols
through.

I haven't worked much on Semantic's smart completion for Emacs Lisp
because the existing tools are very good, so I use those.

Eric

-- 
          Eric Ludlam:                       eric@...
   Siege: http://www.siege-engine.com          Emacs: http://cedet.sourceforge.net

>>> David <de_bb@...> seems to think that:
>"Eric M. Ludlam" <eric@...> writes:
>> I've been trying to find and track these down too.  You can use:
>>
>> M-x semantic-debug-idle-function
>>
>> or
>>
>> M-x semantic-debug-idle-work-function
>>
>> will produce a nice stack.  Thanks.
>
>The idle function seems to choke on highlighting "(require 'info)". But
>maybe this is due to my local Emacs installation... Anyway, the
>backtrace is attached.

I checked in what ought to be a fix for this in
semantic-decorate-include.el.

>> I haven't worked much on Semantic's smart completion for Emacs Lisp
>> because the existing tools are very good, so I use those.
>
>Hmm. I like semantic's idle completion. But I guess this can be done
>with the existing tools, too.

It is certainly fixable in Semantic.  It is likely that fixes for
Emacs Lisp, will be useful for any language that uses the concept of
include files loosely.

>By the way, there is also a bug in the Emacs Lisp completion: if I
>complete a function with s-ia-complete-symbol, an opening bracket gets
>appended, which is not what I want in Emacs Lisp. :-) The idle
>completion works as expected, though.

Ah.  It seems that semantic-ia-insert-tag is not overloadable for
Emacs Lisp.  If this were turned into an overload function, the way
that semantic-up-reference is overloaded, then an Emacs Lisp version
could be written that inserts the right kind of extra text.

I'll add this to my todo list unless someone makes a patch for it
first.

Eric

-- 
          Eric Ludlam:                       eric@...
   Siege: http://www.siege-engine.com          Emacs: http://cedet.sourceforge.net

>>> David <de_bb@...> seems to think that:
>David <de_bb@...> writes:
>> OK, this was easy enough. :-) I attached a patch for semantic-ia.el and
>> semantic-el.el which works for me.
>
>Here's already a new one which is slightly more intelligent by looking
>if the function needs any arguments. Maybe one could also print the
>function arguments in the modeline, but I guess people already use eldoc
>for that. The problem is that eldoc is somehow not triggered by the
>completion - I always have to type another character first.

Thanks for the patch.  I have finally merged it into CVS.  Thanks.

I am not that familiar w/ how eldoc works.  In theory, you can use the
semantic-idle-summary-mode instead of eldoc, but eldoc has much better
output.

>By the way, another thing I noticed: When I want to complete a symbol
>from the Gnus package, I get asked if I want to enable the Gnus
>registry. It seems that somehow gnus-registry-install gets invoked by
>the semantic parsing... But maybe that's more of Gnus bug (I use Gnus
>from CVS)?
  [ ... ]

Semantic will attempt put .el files into a buffer in order to parse
them.  Perhaps gnus files have buffer-local variables that run code in
them?

Eric

-- 
          Eric Ludlam:                       eric@...
   Siege: http://www.siege-engine.com          Emacs: http://cedet.sourceforge.net