CEDET

Hi Eric,

> I have an impending upgrade for `semantic-lex'.
> 
> Many analyzers update the lexer local variable `end-point', and a
> couple others during operation.  This throws warnings during the byte
> compile stage for languages using this technique.  In addition, these
> variables are not documented too well.
> 
> I've put in new variables such as `semantic-lex-end-point', documented
> it, improved the doc strings for related functions, and used that
> during lexical analysis.  Some point soon I'll check in the change for
> a couple of these variables, in addition to updating all languages in
> the `bovine' directory.  Some other languages may cease working after
> this and need updating.
> 
> I hope I can finish this soon.
> 
> Additional thoughts=3F

All that looks good!  Thanks Eric.

As you are working on semantic-lex, I think it would be better to
rename `semantic-lex-token' to `semantic-lex-push-token' as that
function actually push a new token in the lexical stream.  And to
create a true `semantic-lex-token' to create new lexical tokens.
Something like that:

(defmacro semantic-lex-token (symbol start end)
  "Create a lexical token.
SYMBOL is a symbol representing the class of syntax found.
START and END define the bounds of the token in the current buffer."
  `(cons ,symbol (cons ,start ,end)))

(defmacro semantic-lex-push-token (token)
  "Push TOKEN in the lexical analyzer token stream.
Return the updated token stream.
This macro should only be called within the bounds of
`define-lex-analyzer'.  It changes the values of the lexical
analyzer variables `token-stream' and `end-point'.
If you need to move `end-point' somewhere else, just modify this
variable after calling `semantic-lex-push-token'."
  (setq end-point (semantic-lex-token-end ,token)
        token-stream (cons ,token token-stream)))

Thus we will have a true abstraction layer for lexical tokens, like
we have one for semantic tokens.

What do you think=3F
David

P.S.: I checked the fixes in for semantic-el.el and semanticdb-el.el.
Is there any reason for semantic-el.el to be in the bovine
sub-directory=3F  Elisp parser is not the "bovinator", isn't it=3F

I have an impending upgrade for `semantic-lex'.

Many analyzers update the lexer local variable `end-point', and a
couple others during operation.  This throws warnings during the byte
compile stage for languages using this technique.  In addition, these
variables are not documented too well.

I've put in new variables such as `semantic-lex-end-point', documented
it, improved the doc strings for related functions, and used that
during lexical analysis.  Some point soon I'll check in the change for
a couple of these variables, in addition to updating all languages in
the `bovine' directory.  Some other languages may cease working after
this and need updating.

I hope I can finish this soon.

Additional thoughts?
Eric

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org

Hi all,

  I've checked in changes to the new `semantic-debug.el' api,
including the needed changes for the bovinator parser and debugger.

  Unlike the previous debugger, this uses recursive edit, so you can
really snoop about while debugging your parser.

  I thought was doing a pretty good job of keeping things nice and
abstract in semantic-debug.el so it would be easy to build
implementations for wisent right up till I wrote the highlighting
code.  This bit may need some work when trying to get wisent to work.

  Well, I still have to finish my lexical work so I can check in my
copies of the language files I'm working on so you can actually use
it, but if you rebuild a .by file, things should be set up.

Eric

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org

Hi,

  This looks like a refactoring error.  The existence of `filename'
there is erroneous, and works in some cases by accident.  (The caller
is passing in a local variable called `filename' which then scopes in
ok.)

  While I haven't reproduced your error, I'm guessing my recent check
in will fix it.  The item it is filling out is the name of the table,
so it is non-critical, whatever the value may have been.

Eric

*** semanticdb.el.~1.51.~	Mon Feb  3 21:34:52 2003
--- semanticdb.el	Wed Feb 12 20:19:49 2003
***************
*** 150,156 ****
        ;; This implementation will satisfy autoloaded classes
        ;; for tables.
        (setq newtab (funcall (oref db new-table-class)
! 			    (file-name-nondirectory filename)
  			    :file (file-name-nondirectory file)
  			    ))
        (oset newtab parent-db db)
--- 150,156 ----
        ;; This implementation will satisfy autoloaded classes
        ;; for tables.
        (setq newtab (funcall (oref db new-table-class)
! 			    (file-name-nondirectory file)
  			    :file (file-name-nondirectory file)
  			    ))
        (oset newtab parent-db db)


>>> ryk@... seems to think that:
>Hi all,
>
>I started noticing elisp error with semanticdb in the past few days.
>A stack trace is shown at the end of this email.
>I get this when I hit "d e" within the *cvs* buffer created
>by cvs-examine.
>
>It seems like semanticdb-create-table refers to a dynamic variable
>'filename'.  I don't yet understand this and related code, but it
>seems like semanticdb-create-table is relying on one of the callees to
>set 'filename' to a valid value.  An example is find-file.
>
>However cvs-retrieve-revision, which cvs-mode-idiff calls, does not
>set 'filename'.  For your convenience, I've included the defun below.
>If I add one line just below just above (after-find-file) which binds
>'filename' to some valid value, then the error is not signalled.
>
>
>(defun cvs-retrieve-revision (fileinfo rev)
>  "Retrieve the given REVision of the file in FILEINFO into a new buffer."
>  (let* ((file (cvs-fileinfo->full-path fileinfo))
>	 (buffile (concat file "." rev)))
>    (or (find-buffer-visiting buffile)
>	(with-current-buffer (create-file-buffer buffile)
>	  (message "Retrieving revision %s..." rev)
>	  (let ((res (call-process cvs-program nil t nil
>				   "-q" "update" "-p" "-r" rev file)))
>	    (when (and res (not (and (equal 0 res))))
>	      (error "Something went wrong retrieving revision %s: %s" rev res))
>	    (set-buffer-modified-p nil)
>	    (let ((buffer-file-name (expand-file-name file)))
>	      (after-find-file))
>	    (toggle-read-only 1)
>	    (message "Retrieving revision %s... Done" rev)
>	    (current-buffer))))))
>
>
>Debugger entered--Lisp error: (void-variable filename)
>  (file-name-nondirectory filename)
>  (funcall (oref db new-table-class) (file-name-nondirectory filename) :file (file-name-nondirectory file))
  [ ... ]

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org

Thanks David.  These look like great changes.

Eric

>>> "David PONCE" <David.Ponce@...> seems to think that:
>Hi Eric,
>
>Here is a patch for semantic-el.el and semanticdb-el.el that fixes
>various "wrong-type-argument" errors in `senator-jump-interactive'
>and `senator-complete-symbol', in `emacs-lisp-mode'.
>
>In semantic-el.el I fixed override of `semantic-ctxt-current-symbol'
>and `semantic-ctxt-current-function' to ensure they return their
>result in a list as specified in semantic-ctxt.
>
>In semanticdb-el.el I fixed `insert' errors in senator completion,
>when the semantic database is enabled.  The problem was in method
>`semanticdb-find-nonterminal-by-name-regexp-method' that returned nil
>values in the token list, because `semanticdb-elisp-sym->nonterm'
>failed to convert some symbols returned by `apropos-internal' into
>semantic tokens.  The fix ensure that only valid tokens are returned.
>I also simplified the code ;-)
>
>What do you think?
>David
>
>2003-02-12  David Ponce  <david@...>
>
>        * semanticdb-el.el
>        (semanticdb-find-nonterminal-by-name-regexp-method): Ensure it
>        returns valid data.  Code simplified.
>
>        * bovine/semantic-el.el
>        (semantic-ctxt-current-symbol emacs-lisp-mode)
>        (semantic-ctxt-current-function emacs-lisp-mode): Return result
>        in a list.
>	
>Index: semanticdb-el.el
>===================================================================
>RCS file: /cvsroot/cedet/cedet/semantic/semanticdb-el.el,v
>retrieving revision 1.2
>diff -c -r1.2 semanticdb-el.el
>*** semanticdb-el.el	6 Dec 2002 01:33:38 -0000	1.2
>--- semanticdb-el.el	12 Feb 2003 15:53:23 -0000
>***************
>*** 194,221 ****
>  (defmethod semanticdb-find-nonterminal-by-name-regexp-method
>    ((database semanticdb-project-database-emacs-lisp)
>     regex search-parts search-includes diff-mode find-file-match)
>!   "Find all occurances of nonterminals with name matching REGEX in DATABASE.
>! Uses `apropos-internal' to find matches.a
>  See `semanticdb-find-nonterminal-by-function' for details on DATABASES,
>  SEARCH-PARTS, SEARCH-INCLUDES DIFF-MODE, and FIND-FILE-MATCH.
>  Return a list ((DB-TABLE . TOKEN-LIST) ...)."
>!   (let ((syms (apropos-internal regex))
>! 	(toklst nil)
>! 	(newtable nil)
>! 	)
>!     (when syms
>!       (while syms
>! 	(setq toklst (cons (semanticdb-elisp-sym->nonterm (car syms))
>! 			   toklst)
>! 	      syms (cdr syms)))
>!       ;; Set up the table.
>!       (setq newtable (semanticdb-table-emacs-lisp "name regexp search"))
>!       (oset newtable parent-db database)
>!       (oset newtable tokens toklst)
>!       ;; Return the new table.
>!       (list (cons newtable toklst))
>!       )
>!     ))
>  
>  (defmethod semanticdb-find-nonterminal-by-type-method
>    ((database semanticdb-project-database-emacs-lisp)
>--- 194,213 ----
>  (defmethod semanticdb-find-nonterminal-by-name-regexp-method
>    ((database semanticdb-project-database-emacs-lisp)
>     regex search-parts search-includes diff-mode find-file-match)
>!   "Find all occurrences of nonterminals with name matching REGEX in DATABASE.
>! Uses `apropos-internal' to find matches.
>  See `semanticdb-find-nonterminal-by-function' for details on DATABASES,
>  SEARCH-PARTS, SEARCH-INCLUDES DIFF-MODE, and FIND-FILE-MATCH.
>  Return a list ((DB-TABLE . TOKEN-LIST) ...)."
>!   (let ((toklst (delq nil (mapcar 'semanticdb-elisp-sym->nonterm
>! 				  (apropos-internal regex)))))
>!     (if toklst
>! 	;; Set up the table.
>! 	(let ((newtable (semanticdb-table-emacs-lisp "name regexp search")))
>! 	  (oset newtable parent-db database)
>! 	  (oset newtable tokens toklst)
>! 	  ;; Return the new table.
>! 	  (list (cons newtable toklst))))))
>  
>  (defmethod semanticdb-find-nonterminal-by-type-method
>    ((database semanticdb-project-database-emacs-lisp)
>Index: bovine/semantic-el.el
>===================================================================
>RCS file: /cvsroot/cedet/cedet/semantic/bovine/semantic-el.el,v
>retrieving revision 1.6
>diff -c -r1.6 semantic-el.el
>*** bovine/semantic-el.el	29 Jan 2003 03:43:44 -0000	1.6
>--- bovine/semantic-el.el	12 Feb 2003 15:53:23 -0000
>***************
>*** 387,393 ****
>    (save-excursion
>      (if point (goto-char point))
>      (require 'thingatpt)
>!     (thing-at-point 'symbol)))
>  
>  (define-mode-overload-implementation semantic-ctxt-current-assignment emacs-lisp-mode
>    (&optional point)
>--- 387,395 ----
>    (save-excursion
>      (if point (goto-char point))
>      (require 'thingatpt)
>!     (let ((sym (thing-at-point 'symbol)))
>!       (if sym (list sym)))
>!     ))
>  
>  (define-mode-overload-implementation semantic-ctxt-current-assignment emacs-lisp-mode
>    (&optional point)
>***************
>*** 401,407 ****
>    (save-excursion
>      (if point (goto-char point))
>      (semantic-beginning-of-command)
>!     (function-at-point)
>      ))
>  
>  (define-mode-overload-implementation semantic-ctxt-current-argument emacs-lisp-mode
>--- 403,410 ----
>    (save-excursion
>      (if point (goto-char point))
>      (semantic-beginning-of-command)
>!     (let ((fun (function-at-point)))
>!       (if fun (list fun)))
>      ))
>  
>  (define-mode-overload-implementation semantic-ctxt-current-argument emacs-lisp-mode
>
>
>
>
>-------------------------------------------------------
>This SF.NET email is sponsored by:
>SourceForge Enterprise Edition + IBM + LinuxWorld http://www.vasoftware.com
>_______________________________________________
>Cedet-devel mailing list
>Cedet-devel@...
>https://lists.sourceforge.net/lists/listinfo/cedet-devel
>

-- 
          Eric Ludlam:                 zappo@..., eric@...
   Home: http://www.ludlam.net            Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net               GNU: http://www.gnu.org