CEDET

"Eric M. Ludlam" <eric@...> writes:

>   Now in CEDET/CVS is an experimental ctags parsing database backend.
> I have only gotten support for C/C++ so far.

You've got:

(defcustom semantic-ectag-program "~/src/ctags-5.7/ctags" ; "ctags"
  "The Exuberent CTags program to use."
  :group 'semantic
  :type 'file)

Probably "ctags" is the better default!

Also

> (semantic-enable-exuberent-ctags 'c-mode)
> (semantic-enable-exuberent-ctags 'c++-mode)

doesn't work for me.  The function only takes strings.  (Presumably
that's a bug.  Not a significant problem, anyway.)

I get a background error with flymake-mode.  I have flymake-mode
configured so it doesn't operate on filenames with "flymake" in them,
and semantic-inhibit-functions set appropriately to ignore the same
buffers, but I saw errors when semantic was trying to run ctags on
flymake files, so I guess semantic-inhibit-functions isn't being used
properly for the new feature?

[...]

>>> Bruce Stephens <bruce+cedet@...> seems to think that:
>"Eric M. Ludlam" <eric@...> writes:
>
>>   Now in CEDET/CVS is an experimental ctags parsing database backend.
>> I have only gotten support for C/C++ so far.
>
>You've got:
>
>(defcustom semantic-ectag-program "~/src/ctags-5.7/ctags" ; "ctags"
>  "The Exuberent CTags program to use."
>  :group 'semantic
>  :type 'file)
>
>Probably "ctags" is the better default!

Heh, thanks for pointing that out.

>Also
>
>> (semantic-enable-exuberent-ctags 'c-mode)
>> (semantic-enable-exuberent-ctags 'c++-mode)
>
>doesn't work for me.  The function only takes strings.  (Presumably
>that's a bug.  Not a significant problem, anyway.)

Thanks for pointing that out too.  I had also discovered that late
this afternoon, and have checked in a fix.

>I get a background error with flymake-mode.  I have flymake-mode
>configured so it doesn't operate on filenames with "flymake" in them,
>and semantic-inhibit-functions set appropriately to ignore the same
>buffers, but I saw errors when semantic was trying to run ctags on
>flymake files, so I guess semantic-inhibit-functions isn't being used
>properly for the new feature?

Hmmm.  A curious issue.  Those functions are run assuming the current
buffer is the buffer to be parsed.  The ctags support specifically
does not load anything into a buffer, which is the main reason to have
made this extension.  Thus, I can't use those functions.

According to the ctags, you can add --exclude="*flymake*", but I don't
know if that works when a file is explicitly listed, as this tool is
doing.  In any case, you can put that into a ~/.ctags to try it.

Eric

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

Howdy,

  I've changed the ctags installation procedure to be a bit more
streamlined, and checked it into CVS.

  The new configuration is in the various doc files, and there is
a single point for installing all the CTags support.  You can do it
like this in your .emacs file:

(semantic-load-enable-all-exuberent-ctags-support)

  It isn't in the default, because it will throw an error if there is
something wrong with your installed version of ctags.

  Also in the mix, is language support for borne shell scripts,
supplied entirely by ctags.  The sh script prototype was an experiment
to make adding support for a simple language like a shell script via
ctags very easy... It's 2 variables, and a 2 line function.

  So if you are wishing there was support for some language that
happens to be supported by ctags, perhaps now you can add it.

  See cedet/semantic/ctags/semantic-ectag-lang.el for the function
semantic-default-sh-setup.

Eric

>>> "Eric M. Ludlam" <eric@...> seems to think that:
>Hello all,
>
>  Now in CEDET/CVS is an experimental ctags parsing database backend.
>I have only gotten support for C/C++ so far.
  [ ... ]

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

Would it be possible to use etags that comes along with gnu emacs?

/Mats

>>> "Matzi Kratzi" <matzikratzi@...> seems to think that:
>Would it be possible to use etags that comes along with gnu emacs?
  [ ... ]

Hi,

 Etags would be sufficient for finding a single tag by name in a
global flat namespace.  Thus, if you wanted to find a function "foo",
etags could do it.  The next step in smart completion, or showing
function hints is to go to the file the tag was found in, and have
Emacs parse it to get the extra needed information.  The ebrowse
backend to semanticdb does something similar.  Ebrowse does have a
little bit more info I can use though, such as the code
structure. (fields of a struct, etc).

  There are some cute hacks out there that use etags and some
assumptions about how code is laid out to show function hints.
Semantic, however, needs the entire tag broken down into a data
structure which is uses to analyze the code for making suggestions.

 Exuberent ctags includes signature, datatype, and other features.  As
a side effect, I don't need Emacs to parse the file to use the tags,
and that is where the time savings comes in.

Eric

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

"Eric M. Ludlam" <eric@...> writes:

[...]

> According to the ctags, you can add --exclude="*flymake*", but I don't
> know if that works when a file is explicitly listed, as this tool is
> doing.  In any case, you can put that into a ~/.ctags to try it.

Something like this seems to work:

Index: semantic/semantic-idle.el
===================================================================
RCS file: /cvsroot/cedet/cedet/semantic/semantic-idle.el,v
retrieving revision 1.51
diff -u -r1.51 semantic-idle.el
--- semantic/semantic-idle.el	7 Sep 2008 14:16:02 -0000	1.51
+++ semantic/semantic-idle.el	22 Oct 2008 12:30:47 -0000
@@ -463,9 +463,12 @@
       (when (eq (cdr a) major-mode)
 	(dolist (file (directory-files default-directory t (car a) t))
 	  (semantic-throw-on-input 'parsing-mode-buffers)
-	  (save-excursion
-	    (semanticdb-file-table-object file)
-	    ))))
+	  (if (not
+	       (run-hook-with-args-until-success 'semantic-idle-inhibit-functions
+						 file))
+	      (save-excursion
+		(semanticdb-file-table-object file)
+		)))))
     ))
 
 (defun semantic-idle-pnf-test ()
Index: semantic/semantic.el
===================================================================
RCS file: /cvsroot/cedet/cedet/semantic/semantic.el,v
retrieving revision 1.209
diff -u -r1.209 semantic.el
--- semantic/semantic.el	15 Sep 2008 00:24:31 -0000	1.209
+++ semantic/semantic.el	22 Oct 2008 12:30:47 -0000
@@ -261,6 +261,15 @@
   :group 'semantic
   :type 'hook)
 
+(defcustom semantic-idle-inhibit-functions nil
+  "List of functions to call with the name of a file before Semantic tries
+to process it while idle.
+If any of these functions returns non-nil, the file is not processed.  (It may
+be processed if visited, but will not be processed simply because it is
+neighbouring a visited file.)"
+  :group 'semantic
+  :type 'hook)
+
 (defvar semantic-init-hooks nil
   "*Hooks run when a buffer is initialized with a parsing table.")

"Eric M. Ludlam" <eric@...> writes:

>   I've changed the ctags installation procedure to be a bit more
> streamlined, and checked it into CVS.

Cool.

I'm getting an idle error:

	CTAGS/config.h...
	Idle Work Including Error: "#<buffer DsaCSSLConfig.C> - Wrong type argument: stringp, nil"

(I guess the actual filenames aren't significant, except that this is
ctags related.)  config.h is the first file included into
DsaCSSLConfig.C.

I can't seem to debug this.  If I set debug-on-error to t, I don't
seem to get any error (just "CTAGS/config.h...").

[...]

Bruce Stephens <bruce+cedet@...> writes:

[...]

> I'm getting an idle error:
>
> 	CTAGS/config.h...
> 	Idle Work Including Error: "#<buffer DsaCSSLConfig.C> - Wrong type argument: stringp, nil"
>
> (I guess the actual filenames aren't significant, except that this is
> ctags related.)  config.h is the first file included into
> DsaCSSLConfig.C.

My fault.

In c-mode-common-hooks I have a function that conditionally switches
on flymake-mode, and it uses buffer-file-name.  Buffers needn't be
visiting files, of course, and that seems to be what's happening here.

[...]