CEDET

>>> Jake Colman <colman@...> seems to think that:
>
>I am probably being stupid but I'm having trouble getting cedet loaded
>and doing something useful.
>
>I am using cedet-1.0pre4 with XEmacs 21.4.17.
>
>My init.el has he following statements:
>
>(load-file "~/site-lisp/cedet-1.0pre4/common/cedet.el")
>(semantic-load-enable-guady-code-helpers)
>
>Here are some of the issues I am having:
>
>1) Senator/Navigate/Next works.  But the C-c,n accelerator does not.

I am not familiar enough with XEmacs to know why the C-c , keymap
might not work.  Is it listed in the senator menu as the keybinding?

>2) I activated the speedbar class browser.  It fontified a whole mess of
>   files and then displayed a browser bar with a duplicated list of my
>   namespace.  There is nothing in there about my class!

I've just used the class browser on a new batch of C++ code I have
access to, and see that it likely needs some help with namespaces.  I
wrote that before I used namespaces much, so clearly it needs some
help.  I'll add that to my todo list.

Another good bet is to find and use ECB.  It is a great UI that uses
the CEDET data to provide a useful UI allowing structured browsing of
your code.

>3) I can't figure out what/how cedet can do something for me.  I would
>   love to RTFM but neither the Semantic User Guide nor the Semantic
>   Manual tells me how I can use cedet to improve my programming life.
>   Is there something else I should be reading?

Most of the interesting items should be in the menus, so discovery
should be easy.  It appears you picked out something that was broken
to experiment with.

Another common use is the smart completion.  If you want good smart
completion, your best bet is to get the CVS version of CEDET.  It has
a better installation document, and much better smart completion than
pre4.

>4) I used to use the oo-browser.  I have also used Eclipse just a little
>   bit to see what my colleagues were raving about.  I'm looking for
>   something that allows me to easily browse the classes that make up my
>   project, that allows me to jump to, and/or display, a method
>   definition from where it's being used, etc.  Can cedet do this kind
>   of stuff for me?
  [ ... ]

CEDET has the infrastructure to show you method definitions, and also
smart completion.  Mixing CEDET and ECB will give you the rest.  ECB
is at http://ecb.sf.net.

The CVS version of CEDET has more infrastructure needed to make setup
a little more understandable, and for large projects, it has better
ways of configuring Emacs to find files in a project.

Eric

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

On Tue, Apr 29, 2008 at 6:42 AM, Jake Colman <colman@...> wrote:
>
...
>  3) I can't figure out what/how cedet can do something for me.  I would
>    love to RTFM but neither the Semantic User Guide nor the Semantic
>    Manual tells me how I can use cedet to improve my programming life.
>    Is there something else I should be reading?

Those of you who find CEDET helpful can help us who have recently
installed by adding your tips to the wiki page at
http://www.emacswiki.org/cgi-bin/wiki/CollectionOfEmacsDevelopmentEnvironmentTools
If there's something in Emacs I can't figure out from its own help,
emacswiki is where I go next. The cedet page there is rather sparse,
it could use "How it can help you" and "Examples" sections. Even links
to blog posts and messages in list archives would be good, collected
in one place...

>>> Jake Colman <colman@...> seems to think that:
>>>>>> "EML" == Eric M Ludlam <eric@...> writes:
  [ ... ]
>   >> 2) I activated the speedbar class browser.  It fontified a whole mess of
>   >> files and then displayed a browser bar with a duplicated list of my
>   >> namespace.  There is nothing in there about my class!
>
>   EML> I've just used the class browser on a new batch of C++ code I have
>   EML> access to, and see that it likely needs some help with namespaces.  I
>   EML> wrote that before I used namespaces much, so clearly it needs some
>   EML> help.  I'll add that to my todo list.
>
>Thanks.  I guess you'll post availability here.

Yes, I post to cedet-devel when I check in big changes.

>   EML> Another good bet is to find and use ECB.  It is a great UI that uses
>   EML> the CEDET data to provide a useful UI allowing structured browsing of
>   EML> your code.
>
>ECB is also one of your packages, correct?  Is it intended to complement
>CEDET?  I thought CEDET was intended to be the superset of everything.
>Is it your intention for power programmers to install both CEDET and
>ECB?  Does ECB document its usage of CEDET?

ECB was written by others and happens to depend on CEDET for some of
it's functionality.

>   >> 4) I used to use the oo-browser.  I have also used Eclipse just a little
>   >> bit to see what my colleagues were raving about.  I'm looking for
>   >> something that allows me to easily browse the classes that make up my
>   >> project, that allows me to jump to, and/or display, a method
>   >> definition from where it's being used, etc.  Can cedet do this kind
>   >> of stuff for me?
>   EML>   [ ... ]
>
>   EML> CEDET has the infrastructure to show you method definitions, and also
>   EML> smart completion.  Mixing CEDET and ECB will give you the rest.  ECB
>   EML> is at http://ecb.sf.net.
>
>   EML> The CVS version of CEDET has more infrastructure needed to make setup
>   EML> a little more understandable, and for large projects, it has better
>   EML> ways of configuring Emacs to find files in a project.
>
>I am having trouble accessing the cvs repository from my work machine.
>Is it possible to get access to a tar file containing the latest version
>in CVS?
  [ ... ]

I'm in the middle of testing for producing another pre-release, so you
may just need to wait a few days.   I don't have the infrastructure to
create nightly builds for CEDET.

Eric

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

>>> yary <not.com@...> seems to think that:
>Hello-
>
>I'm using the CVS version of cedet. With senator's sticky func mode
>on, clicking on the header lines gives a menu allowing one to copy the
>tag and a few other things.
>
>The first option, senator-copy-tag, seemed to be broken, then I
>realized it copies the tag to a tag ring, not the kill ring. This
>seems not so useful, since when I copy something I usually want to
>yank it soon after, regardless of how I copied/killed it. On the other
>hand I assume there's a reason for the existence of a tag ring that's
>distinct from a kill ring.
>
>So, what's the reasoning and usage of the tag ring? How can it help me?
  [ ... ]

The tag ring is useful for inserting those tags back into code
elsewhere.  You can copy a tag from one place, and if you insert the
tag with senator's insert tag function, it will stick in, say, a
function call to that tag, or a prototype of the tag.  If you paste
the tag into a buffer of a different mode, such as texinfo, it will
insert it as documentation.

If you "kill" a tag, it will stick it in the tag ring, and put the
text into the regular kill ring for pasting later.

If you want to copy the text of a tag into the kill ring, you can use
the regular `mark-defun' (C-M-h) and `kill-ring-save' (M-w) and
Senator uses advice to make it work correctly.

If there is some combo feature like this you think would be useful to
add to the menu, or some other interface, let me know, or better yet,
provide a patch.

Thanks
Eric

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

Hi,

  It did not occur to me that someone would want the pre-processor
completely disabled.  Mostly for cases like this:

#ifdef FOO
int someFcn(int a)
#else
int otherFcn(int b)
#endif
{
  // ...
}

In your case though, you could just define all
those variables so they enable all those blocks of code.  You can
customize either of these variables:

semantic-lex-c-preprocessor-symbol-map
semantic-lex-c-preprocessor-symbol-file

The first is useful if you have some very simple entries.  The
second is new, and makes it easy to just write a C file to define your
stuff, and have that table included in all your sources.

Most of the hacks I can think of would still defer #else or !define
sections.

Good Luck
Eric

>>> Rob Walker <rob.lists@...> seems to think that:
>Hi,
>
>I've just updated my semantic install to the cvs version and am very impressed 
>with the speed improvements since my last update (about a year ago).
>
>However, I notice semantic is now processing the C preprocessor statements and 
>selectively including or excluding code depending on the values assigned to 
>each macro.  Is there any way to disable this?  I have to work on some code 
>that has lots of code like this:
>
>#if defined(FOO)
>void functionForDeviceFoo()
>{
>  // ....
>}
>#endif
>
>#if defined(BAR)
>void functionForDeviceBar()
>{
>  // ....
>}
>#endif
>
>....
>
>
>The same file is then built several times for different devices (Don't ask - 
>it's legacy code).  I need to be able to jump between all the different 
>functions without having to say each time whether FOO or BAR is defined, so 
>I'd like to turn semantic's preprocessing off.  Can this be done?
  [ ... ]

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

Eric,

So there's no way to disable it for cases like?:

#if defined(FOO)
void fooFunc()
{
}

#else

void barFunc()
{
}

#endif

Thanks

Rob

On Saturday 19 April 2008, Eric M. Ludlam wrote:
> Hi,
>
>   It did not occur to me that someone would want the pre-processor
> completely disabled.  Mostly for cases like this:
>
> #ifdef FOO
> int someFcn(int a)
> #else
> int otherFcn(int b)
> #endif
> {
>   // ...
> }
>
> In your case though, you could just define all
> those variables so they enable all those blocks of code.  You can
> customize either of these variables:
>
> semantic-lex-c-preprocessor-symbol-map
> semantic-lex-c-preprocessor-symbol-file
>
> The first is useful if you have some very simple entries.  The
> second is new, and makes it easy to just write a C file to define your
> stuff, and have that table included in all your sources.
>
> Most of the hacks I can think of would still defer #else or !define
> sections.
>
> Good Luck
> Eric
>
> >>> Rob Walker <rob.lists@...> seems to think that:
> >
> >Hi,
> >
> >I've just updated my semantic install to the cvs version and am very
> > impressed with the speed improvements since my last update (about a year
> > ago).
> >
> >However, I notice semantic is now processing the C preprocessor statements
> > and selectively including or excluding code depending on the values
> > assigned to each macro.  Is there any way to disable this?  I have to
> > work on some code that has lots of code like this:
> >
> >#if defined(FOO)
> >void functionForDeviceFoo()
> >{
> >  // ....
> >}
> >#endif
> >
> >#if defined(BAR)
> >void functionForDeviceBar()
> >{
> >  // ....
> >}
> >#endif
> >
> >....
> >
> >
> >The same file is then built several times for different devices (Don't ask
> > - it's legacy code).  I need to be able to jump between all the different
> > functions without having to say each time whether FOO or BAR is defined,
> > so I'd like to turn semantic's preprocessing off.  Can this be done?
>
>   [ ... ]



-- 
Rob Walker <rob@...>
http://www.tenfoot.org.uk

Hi,

  I checked in a change to CVS to add
`semantic-c-obey-conditional-section-parsing-flag' which ought to do
what you want.  This could make parsing system header files a bit
problematic.  In that case, you may need to use it as a local
variable, or project local variable.

Good Luck
Eric

>>> Rob Walker <rob.lists@...> seems to think that:
>Eric,
>
>So there's no way to disable it for cases like?:
>
>#if defined(FOO)
>void fooFunc()
>{
>}
>
>#else
>
>void barFunc()
>{
>}
>
>#endif
>
>Thanks
>
>Rob
>
>On Saturday 19 April 2008, Eric M. Ludlam wrote:
>> Hi,
>>
>>   It did not occur to me that someone would want the pre-processor
>> completely disabled.  Mostly for cases like this:
>>
>> #ifdef FOO
>> int someFcn(int a)
>> #else
>> int otherFcn(int b)
>> #endif
>> {
>>   // ...
>> }
>>
>> In your case though, you could just define all
>> those variables so they enable all those blocks of code.  You can
>> customize either of these variables:
>>
>> semantic-lex-c-preprocessor-symbol-map
>> semantic-lex-c-preprocessor-symbol-file
>>
>> The first is useful if you have some very simple entries.  The
>> second is new, and makes it easy to just write a C file to define your
>> stuff, and have that table included in all your sources.
>>
>> Most of the hacks I can think of would still defer #else or !define
>> sections.
>>
>> Good Luck
>> Eric
>>
>> >>> Rob Walker <rob.lists@...> seems to think that:
>> >
>> >Hi,
>> >
>> >I've just updated my semantic install to the cvs version and am very
>> > impressed with the speed improvements since my last update (about a year
>> > ago).
>> >
>> >However, I notice semantic is now processing the C preprocessor statements
>> > and selectively including or excluding code depending on the values
>> > assigned to each macro.  Is there any way to disable this?  I have to
>> > work on some code that has lots of code like this:
>> >
>> >#if defined(FOO)
>> >void functionForDeviceFoo()
>> >{
>> >  // ....
>> >}
>> >#endif
>> >
>> >#if defined(BAR)
>> >void functionForDeviceBar()
>> >{
>> >  // ....
>> >}
>> >#endif
>> >
>> >....
>> >
>> >
>> >The same file is then built several times for different devices (Don't ask
>> > - it's legacy code).  I need to be able to jump between all the different
>> > functions without having to say each time whether FOO or BAR is defined,
>> > so I'd like to turn semantic's preprocessing off.  Can this be done?
>>
>>   [ ... ]
>
>
>

Eric,

Thanks for making that change.  I've set it buffer local for the files within 
my project and its working well.

Thanks

Rob


On Sunday 20 April 2008, Eric M. Ludlam wrote:
> Hi,
>
>   I checked in a change to CVS to add
> `semantic-c-obey-conditional-section-parsing-flag' which ought to do
> what you want.  This could make parsing system header files a bit
> problematic.  In that case, you may need to use it as a local
> variable, or project local variable.
>
> Good Luck
> Eric
>
> >>> Rob Walker <rob.lists@...> seems to think that:
> >
> >Eric,
> >
> >So there's no way to disable it for cases like?:
> >
> >#if defined(FOO)
> >void fooFunc()
> >{
> >}
> >
> >#else
> >
> >void barFunc()
> >{
> >}
> >
> >#endif
> >
> >Thanks
> >
> >Rob
> >
> >On Saturday 19 April 2008, Eric M. Ludlam wrote:
> >> Hi,
> >>
> >>   It did not occur to me that someone would want the pre-processor
> >> completely disabled.  Mostly for cases like this:
> >>
> >> #ifdef FOO
> >> int someFcn(int a)
> >> #else
> >> int otherFcn(int b)
> >> #endif
> >> {
> >>   // ...
> >> }
> >>
> >> In your case though, you could just define all
> >> those variables so they enable all those blocks of code.  You can
> >> customize either of these variables:
> >>
> >> semantic-lex-c-preprocessor-symbol-map
> >> semantic-lex-c-preprocessor-symbol-file
> >>
> >> The first is useful if you have some very simple entries.  The
> >> second is new, and makes it easy to just write a C file to define your
> >> stuff, and have that table included in all your sources.
> >>
> >> Most of the hacks I can think of would still defer #else or !define
> >> sections.
> >>
> >> Good Luck
> >> Eric
> >>
> >> >>> Rob Walker <rob.lists@...> seems to think that:
> >> >
> >> >Hi,
> >> >
> >> >I've just updated my semantic install to the cvs version and am very
> >> > impressed with the speed improvements since my last update (about a
> >> > year ago).
> >> >
> >> >However, I notice semantic is now processing the C preprocessor
> >> > statements and selectively including or excluding code depending on
> >> > the values assigned to each macro.  Is there any way to disable this? 
> >> > I have to work on some code that has lots of code like this:
> >> >
> >> >#if defined(FOO)
> >> >void functionForDeviceFoo()
> >> >{
> >> >  // ....
> >> >}
> >> >#endif
> >> >
> >> >#if defined(BAR)
> >> >void functionForDeviceBar()
> >> >{
> >> >  // ....
> >> >}
> >> >#endif
> >> >
> >> >....
> >> >
> >> >
> >> >The same file is then built several times for different devices (Don't
> >> > ask - it's legacy code).  I need to be able to jump between all the
> >> > different functions without having to say each time whether FOO or BAR
> >> > is defined, so I'd like to turn semantic's preprocessing off.  Can
> >> > this be done?
> >>
> >>   [ ... ]

Thanks for the patch.  I hadn't known how to fix that flickering
problem, so eventually switched to some different displayor types.

Your patch did fix the flickering problem.  It also positioned the
tooltip way off on the left.  I keep my Emacs frame on the right side
of my monitor, so left was not offset by the location of the active
frame.  This was easy to fix.  The top-offset is a bit more difficult
since I need to guess the height of the tip.  This is a bit I'm not to
fond of in my solution before.  If anyone has a good idea for solving
this let me know.

I appended my updated function below.

Thanks
Eric

>>> Dan Kruchinin <dan.kruchinin@...> seems to think that:
>Hello.
>
>semantic-displayor-tooltip class has several problems and one of them is 
>critical.
>After setting semantic-complete-inline-analyzer-displayor-class to 
>'semantic-displayor-tooltip
>and trying to make semantic-complete-analyze-inline on some string or 
>symbol, semantic displays
>tooltip with possible completions and sometimes *immediately* hides 
>it(tooltip remains alive for several milliseconds).
>
>This occurs because of tooltip-mode by default enables hook which 
>destroys tooltip when mouse lies on or moves throw it.
>So when semantic wants to display tooltip with completions it at first 
>moves mouse at cursor point, then displays tooltip and moves
>mouse at it's previous position. Sometimes this "previous position" lies 
>on tooltip and sometimes mouse cursor has to move throw tooltiop
>to return back. In both these cases tooltip will be destroyed.
>
>The following patch fixes this problem also with the approach using in 
>patch, user wouldn't see how mouse cursor moves
>(by default cursor moves quite slow and this movement is noticeable for 
>human eye):
  [ ... ]


(defun semantic-displayor-tooltip-show (text)
  "Display a tooltip with TEXT near cursor."
  (let* ((saved-pos (mouse-position))
	 (point-pos (semantic-displayor-point-position))
	 (point-pix-pos nil)
	 (left (frame-parameter (selected-frame) 'left))
	 (top (frame-parameter (selected-frame) 'top))
	 (height (frame-parameter (selected-frame) 'height))
	 (hpix (frame-pixel-height))
	 (linepix (/ hpix height))
	 (numline (length (split-string text "\n")))
	 (newh (* numline linepix))
	 )
    (set-mouse-position (selected-frame) (car point-pos) (cdr point-pos))
    (setq point-pix-pos (mouse-pixel-position))
    (set-mouse-position (car saved-pos) (cadr saved-pos) (cddr saved-pos))
    (let ((tooltip-frame-parameters (append tooltip-frame-parameters nil)))
      (push
       (cons 'left (+ (cadr point-pix-pos) left)) tooltip-frame-parameters)
      (push
       (cons 'top (+ (cddr point-pix-pos) top (- newh))) tooltip-frame-parameters)
      (tooltip-show text)
      )))

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

Eric M. Ludlam wrote:
> Thanks for the patch.  I hadn't known how to fix that flickering
> problem, so eventually switched to some different displayor types.
>
> Your patch did fix the flickering problem.  It also positioned the
> tooltip way off on the left.  I keep my Emacs frame on the right side
> of my monitor, so left was not offset by the location of the active
> frame.  This was easy to fix.  The top-offset is a bit more difficult
> since I need to guess the height of the tip.  This is a bit I'm not to
> fond of in my solution before.  If anyone has a good idea for solving
> this let me know.
>
> I appended my updated function below.
>
> Thanks
> Eric
>
>   

Yes, you're right, I havn't keep attention to such situations.
I wrote new patch which should fix all these problems.
It doesn't touch mouse any more and also it displays tooltip
at the a bit lower position than cursor's one.
You can see the patch at the end of this message.

W.B.R.
Dan Kruchinin

>>>> Dan Kruchinin <dan.kruchinin@...> seems to think that:
>>>>         
>> Hello.
>>
>> semantic-displayor-tooltip class has several problems and one of them is 
>> critical.
>> After setting semantic-complete-inline-analyzer-displayor-class to 
>> 'semantic-displayor-tooltip
>> and trying to make semantic-complete-analyze-inline on some string or 
>> symbol, semantic displays
>> tooltip with possible completions and sometimes *immediately* hides 
>> it(tooltip remains alive for several milliseconds).
>>
>> [...]
>>     
>   [ ... ]
>
>
> (defun semantic-displayor-tooltip-show (text)
>   "Display a tooltip with TEXT near cursor."
>   (let* ((saved-pos (mouse-position))
> 	 (point-pos (semantic-displayor-point-position))
> 	 (point-pix-pos nil)
> 	 (left (frame-parameter (selected-frame) 'left))
> 	 (top (frame-parameter (selected-frame) 'top))
> 	 (height (frame-parameter (selected-frame) 'height))
> 	 (hpix (frame-pixel-height))
> 	 (linepix (/ hpix height))
> 	 (numline (length (split-string text "\n")))
> 	 (newh (* numline linepix))
> 	 )
>     (set-mouse-position (selected-frame) (car point-pos) (cdr point-pos))
>     (setq point-pix-pos (mouse-pixel-position))
>     (set-mouse-position (car saved-pos) (cadr saved-pos) (cddr saved-pos))
>     (let ((tooltip-frame-parameters (append tooltip-frame-parameters nil)))
>       (push
>        (cons 'left (+ (cadr point-pix-pos) left)) tooltip-frame-parameters)
>       (push
>        (cons 'top (+ (cddr point-pix-pos) top (- newh))) tooltip-frame-parameters)
>       (tooltip-show text)
>       )))
>
>   

### the patch for cedet/semantic/semantic-complete.el, rev #1.51###
--- cedet/semantic/semantic-complete.el    2008-04-22 20:30:02.000000000 
+0400
+++ cedet/semantic/semantic-complete.el.patched    2008-04-22 
20:32:46.000000000 +0400
@@ -1624,32 +1624,28 @@
     ))
 
 (defun semantic-displayor-point-position ()
-  "Return the location of POINT as positioned on the selected frame.
+  "Return the location of POINT on screen in pixels.
 Return a cons cell (X . Y)"
-  (let* ((w (selected-window))
-     (f (selected-frame))
-     (edges (semantic-displayor-window-edges w))
-     (col (current-column))
-     (row (count-lines (window-start w) (point)))
-     (x (+ (car edges) col))
-     (y (+ (car (cdr edges)) row)))
-    (cons x y))
-  )
+  (let* ((frame (selected-frame))
+     (left (frame-parameter frame 'left))
+     (top (frame-parameter frame 'top))
+     (point-pix-pos (posn-x-y (posn-at-point)))
+     (edges (window-inside-pixel-edges (selected-window))))
+    (cons (+ (car point-pix-pos) (car edges) left)
+          (+ (cdr point-pix-pos) (cadr edges) top))))
 
 (defun semantic-displayor-tooltip-show (text)
   "Display a tooltip with TEXT near cursor."
-  (let* ((P (semantic-displayor-point-position))
-     (frame (selected-frame))
-     (x (car P))
-     (y (cdr P))
-     (oP (mouse-pixel-position))
-     (tooltip-x-offset 0)
-     (tooltip-y-offset -40)
-     )
-    (set-mouse-position frame x y)
-    (tooltip-show text)
-    (set-mouse-pixel-position (nth 0 oP) (nth 1 oP) (nthcdr 2 oP))
-    ))
+  (let ((point-pix-pos (semantic-displayor-point-position))
+     (tooltip-frame-parameters
+      (append tooltip-frame-parameters nil)))
+    (push
+     (cons 'left (+ (car point-pix-pos) (frame-char-width)))
+     tooltip-frame-parameters)
+    (push
+     (cons 'top (+ (cdr point-pix-pos) (frame-char-height)))
+     tooltip-frame-parameters)
+    (tooltip-show text)))
 
 (defmethod semantic-displayor-scroll-request ((obj 
semantic-displayor-tooltip))
   "A request to for the displayor to scroll the completion list (if 
needed)."
### end ###

Hm, It seems neither of all these patches doesn't really fix all problems.
When your mouse cursor is inside only one frame all is ok but when you're
trying to move in to other frame or just move it on any emacs link, tooltip
with completions disappears. I just looked at  the emacs C code which 
provides
basic API for tooltips. It supposes only one tooltip can be active at 
particular moment.
Therefor tooltip with completions dies when mouse cursor is moved to any 
area which
providing another tooltip(for example, emacs link or mode-line). I 
haven't find any
clear or even "dirty hack" way to solve this problem. Unfortunately 
tooltips were designed
for other purposes.
> Eric M. Ludlam wrote:
> [...]
>>
>>
>>
>> (defun semantic-displayor-tooltip-show (text)
>>   "Display a tooltip with TEXT near cursor."
>>   (let* ((saved-pos (mouse-position))
>>      (point-pos (semantic-displayor-point-position))
>>      (point-pix-pos nil)
>>      (left (frame-parameter (selected-frame) 'left))
>>      (top (frame-parameter (selected-frame) 'top))
>>      (height (frame-parameter (selected-frame) 'height))
>>      (hpix (frame-pixel-height))
>>      (linepix (/ hpix height))
>>      (numline (length (split-string text "\n")))
>>      (newh (* numline linepix))
>>      )
>>     (set-mouse-position (selected-frame) (car point-pos) (cdr 
>> point-pos))
>>     (setq point-pix-pos (mouse-pixel-position))
>>     (set-mouse-position (car saved-pos) (cadr saved-pos) (cddr 
>> saved-pos))
>>     (let ((tooltip-frame-parameters (append tooltip-frame-parameters 
>> nil)))
>>       (push
>>        (cons 'left (+ (cadr point-pix-pos) left)) 
>> tooltip-frame-parameters)
>>       (push
>>        (cons 'top (+ (cddr point-pix-pos) top (- newh))) 
>> tooltip-frame-parameters)
>>       (tooltip-show text)
>>       )))
>>
>>   
>
> ### the patch for cedet/semantic/semantic-complete.el, rev #1.51###
> --- cedet/semantic/semantic-complete.el    2008-04-22 
> 20:30:02.000000000 +0400
> +++ cedet/semantic/semantic-complete.el.patched    2008-04-22 
> 20:32:46.000000000 +0400
> @@ -1624,32 +1624,28 @@
>     ))
>
> (defun semantic-displayor-point-position ()
> -  "Return the location of POINT as positioned on the selected frame.
> +  "Return the location of POINT on screen in pixels.
> Return a cons cell (X . Y)"
> -  (let* ((w (selected-window))
> -     (f (selected-frame))
> -     (edges (semantic-displayor-window-edges w))
> -     (col (current-column))
> -     (row (count-lines (window-start w) (point)))
> -     (x (+ (car edges) col))
> -     (y (+ (car (cdr edges)) row)))
> -    (cons x y))
> -  )
> +  (let* ((frame (selected-frame))
> +     (left (frame-parameter frame 'left))
> +     (top (frame-parameter frame 'top))
> +     (point-pix-pos (posn-x-y (posn-at-point)))
> +     (edges (window-inside-pixel-edges (selected-window))))
> +    (cons (+ (car point-pix-pos) (car edges) left)
> +          (+ (cdr point-pix-pos) (cadr edges) top))))
>
> (defun semantic-displayor-tooltip-show (text)
>   "Display a tooltip with TEXT near cursor."
> -  (let* ((P (semantic-displayor-point-position))
> -     (frame (selected-frame))
> -     (x (car P))
> -     (y (cdr P))
> -     (oP (mouse-pixel-position))
> -     (tooltip-x-offset 0)
> -     (tooltip-y-offset -40)
> -     )
> -    (set-mouse-position frame x y)
> -    (tooltip-show text)
> -    (set-mouse-pixel-position (nth 0 oP) (nth 1 oP) (nthcdr 2 oP))
> -    ))
> +  (let ((point-pix-pos (semantic-displayor-point-position))
> +     (tooltip-frame-parameters
> +      (append tooltip-frame-parameters nil)))
> +    (push
> +     (cons 'left (+ (car point-pix-pos) (frame-char-width)))
> +     tooltip-frame-parameters)
> +    (push
> +     (cons 'top (+ (cdr point-pix-pos) (frame-char-height)))
> +     tooltip-frame-parameters)
> +    (tooltip-show text)))
>
> (defmethod semantic-displayor-scroll-request ((obj 
> semantic-displayor-tooltip))
>   "A request to for the displayor to scroll the completion list (if 
> needed)."
> ### end ###

Hi,

  The tooltip is quite hard to use this way.  I've tried menus too.
On MS windows the menu is nice because you can use the cursor keys to
select stuff from it.  company-mode, which I found on emacswiki, has
some sort of neat text tip it uses.  I thought it might be possible to
use a full-up frame too, but I don't know how to make the window
manager decorations go away.

  I checked in some recent version from here which seemed to work ok
for me.  Perhaps if CEDET gets merged into Emacs as is currently being
discussed on emacs-devel, it will cause Emacs to get fixed.

Eric

>>> Dan Kruchinin <dan.kruchinin@...> seems to think that:
>Hm, It seems neither of all these patches doesn't really fix all problems.
>When your mouse cursor is inside only one frame all is ok but when you're
>trying to move in to other frame or just move it on any emacs link, tooltip
>with completions disappears. I just looked at  the emacs C code which 
>provides
>basic API for tooltips. It supposes only one tooltip can be active at 
>particular moment.
>Therefor tooltip with completions dies when mouse cursor is moved to any 
>area which
>providing another tooltip(for example, emacs link or mode-line). I 
>haven't find any
>clear or even "dirty hack" way to solve this problem. Unfortunately 
>tooltips were designed
>for other purposes.
>> Eric M. Ludlam wrote:
>> [...]
>>>
>>>
>>>
>>> (defun semantic-displayor-tooltip-show (text)
>>>   "Display a tooltip with TEXT near cursor."
>>>   (let* ((saved-pos (mouse-position))
>>>      (point-pos (semantic-displayor-point-position))
>>>      (point-pix-pos nil)
>>>      (left (frame-parameter (selected-frame) 'left))
>>>      (top (frame-parameter (selected-frame) 'top))
>>>      (height (frame-parameter (selected-frame) 'height))
>>>      (hpix (frame-pixel-height))
>>>      (linepix (/ hpix height))
>>>      (numline (length (split-string text "\n")))
>>>      (newh (* numline linepix))
>>>      )
>>>     (set-mouse-position (selected-frame) (car point-pos) (cdr 
>>> point-pos))
>>>     (setq point-pix-pos (mouse-pixel-position))
>>>     (set-mouse-position (car saved-pos) (cadr saved-pos) (cddr 
>>> saved-pos))
>>>     (let ((tooltip-frame-parameters (append tooltip-frame-parameters 
>>> nil)))
>>>       (push
>>>        (cons 'left (+ (cadr point-pix-pos) left)) 
>>> tooltip-frame-parameters)
>>>       (push
>>>        (cons 'top (+ (cddr point-pix-pos) top (- newh))) 
>>> tooltip-frame-parameters)
>>>       (tooltip-show text)
>>>       )))
>>>
>>>   
>>
>> ### the patch for cedet/semantic/semantic-complete.el, rev #1.51###
>> --- cedet/semantic/semantic-complete.el    2008-04-22 
>> 20:30:02.000000000 +0400
>> +++ cedet/semantic/semantic-complete.el.patched    2008-04-22 
>> 20:32:46.000000000 +0400
>> @@ -1624,32 +1624,28 @@
>>     ))
>>
>> (defun semantic-displayor-point-position ()
>> -  "Return the location of POINT as positioned on the selected frame.
>> +  "Return the location of POINT on screen in pixels.
>> Return a cons cell (X . Y)"
>> -  (let* ((w (selected-window))
>> -     (f (selected-frame))
>> -     (edges (semantic-displayor-window-edges w))
>> -     (col (current-column))
>> -     (row (count-lines (window-start w) (point)))
>> -     (x (+ (car edges) col))
>> -     (y (+ (car (cdr edges)) row)))
>> -    (cons x y))
>> -  )
>> +  (let* ((frame (selected-frame))
>> +     (left (frame-parameter frame 'left))
>> +     (top (frame-parameter frame 'top))
>> +     (point-pix-pos (posn-x-y (posn-at-point)))
>> +     (edges (window-inside-pixel-edges (selected-window))))
>> +    (cons (+ (car point-pix-pos) (car edges) left)
>> +          (+ (cdr point-pix-pos) (cadr edges) top))))
>>
>> (defun semantic-displayor-tooltip-show (text)
>>   "Display a tooltip with TEXT near cursor."
>> -  (let* ((P (semantic-displayor-point-position))
>> -     (frame (selected-frame))
>> -     (x (car P))
>> -     (y (cdr P))
>> -     (oP (mouse-pixel-position))
>> -     (tooltip-x-offset 0)
>> -     (tooltip-y-offset -40)
>> -     )
>> -    (set-mouse-position frame x y)
>> -    (tooltip-show text)
>> -    (set-mouse-pixel-position (nth 0 oP) (nth 1 oP) (nthcdr 2 oP))
>> -    ))
>> +  (let ((point-pix-pos (semantic-displayor-point-position))
>> +     (tooltip-frame-parameters
>> +      (append tooltip-frame-parameters nil)))
>> +    (push
>> +     (cons 'left (+ (car point-pix-pos) (frame-char-width)))
>> +     tooltip-frame-parameters)
>> +    (push
>> +     (cons 'top (+ (cdr point-pix-pos) (frame-char-height)))
>> +     tooltip-frame-parameters)
>> +    (tooltip-show text)))
>>
>> (defmethod semantic-displayor-scroll-request ((obj 
>> semantic-displayor-tooltip))
>>   "A request to for the displayor to scroll the completion list (if 
>> needed)."
>> ### end ###
>

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

Hi Yari,

You might also be interested in the fairly new yasnippet for textmate
like completion/expansion of template, see:

   http://code.google.com/p/yasnippet/

Cheers,

On Wed, Apr 23, 2008 at 7:44 PM, yary <not.com@...> wrote:
> I'm just installing CEDET and was a bit dismayed to have to dig a few
>  levels deep in the info files to see how to get code completion
>  working. It's one of the "sexy" features of semantic & CEDET and such
>  should be easy to get going out of the box.
>
>  How about adding something like the following lines to CEDETs install tips:
>
>  ;; Bind semantic code completion to useful keys
>
>  ;; Uncomment next line to have "." start autocompletion in C++ mode
>  ;; (define-key c-mode-map "." 'semantic-complete-self-insert)
>  ;; Uncomment next line to cycle through autocompletions on ctrl-tab
>  ;; (define-key c-mode-map (kbd "C-<tab>") 'semantic-ia-complete-symbol-menu)
>
>  I say 'something like' because I'm a C programmer so am not sure if
>  there's a more specific C++ mode-map, and maybe there should be Java,
>  javascript, lisp, perl, python, php, pascal, ... examples too. Also
>  I'm not sure what elisp-function if any cycles through autocompletions
>  (though with symbol-menu, it's easy to figure out which keystrokes
>  select).
>
>  Also, I thought there was a way to expand common syntax constructs, eg
>  after typing "if" have the buffer insert "if () {}" with the point
>  inside the parens. Is that in CEDET? If so, how does one enable that,
>  and can it be added in the install help?
>
>  -------------------------------------------------------------------------
>  This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
>  Don't miss this year's exciting event. There's still time to save $100.
>  Use priority code J8TL2D2.
>  http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
>  _______________________________________________
>  Cedet-devel mailing list
>  Cedet-devel@...
>  https://lists.sourceforge.net/lists/listinfo/cedet-devel
>



-- 
Bas
Sent from my Espressomachine

>>> yary <not.com@...> seems to think that:
>I'm just installing CEDET and was a bit dismayed to have to dig a few
>levels deep in the info files to see how to get code completion
>working. It's one of the "sexy" features of semantic & CEDET and such
>should be easy to get going out of the box.
>
>How about adding something like the following lines to CEDETs install tips:
>
>;; Bind semantic code completion to useful keys
>
>;; Uncomment next line to have "." start autocompletion in C++ mode
>;; (define-key c-mode-map "." 'semantic-complete-self-insert)
>;; Uncomment next line to cycle through autocompletions on ctrl-tab
>;; (define-key c-mode-map (kbd "C-<tab>") 'semantic-ia-complete-symbol-menu)
>
>I say 'something like' because I'm a C programmer so am not sure if
>there's a more specific C++ mode-map, and maybe there should be Java,
>javascript, lisp, perl, python, php, pascal, ... examples too. Also
>I'm not sure what elisp-function if any cycles through autocompletions
>(though with symbol-menu, it's easy to figure out which keystrokes
>select).

Hi,

  That seems like a good idea.  There is a place for info like that in
the cedet install manual, so I'll update the doc to include more of
the completion options.

>Also, I thought there was a way to expand common syntax constructs, eg
>after typing "if" have the buffer insert "if () {}" with the point
>inside the parens. Is that in CEDET? If so, how does one enable that,
>and can it be added in the install help?
  [ ... ]

CEDET (the CVS version) includes SRecode which is a template library
that could do that, but that UI is not there for it, and SRecode
is overkill for that kind of task.  Emacs comes with abbrev, and
someone recommended yasnippet.  See emacswiki.org's template category
for the many tools that can do that.

Eric

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

On date Thursday 2008-04-10 10:31:23 -0400, Eric M. Ludlam wrote:
> Hi Stefano - 
> 
> I apologize for this very late email.  I am finding lots of mail from
> last summer I did not respond to.  I'm trying to get through all my
> lost email and found yours.
> 
> >>> Stefano Sabatini <stefano.sabatini-lala@...> seems to think that:
> >Hi Eric and company,
> >
> >I'm a great fan of cedet and ECB from more than one year now (and
> >still trying to understand how many things work... ;-)).
> >
> >This patch adds support in the Makefile for an install-info and
> >uninstall-info rule.
> 
> 
> Thanks for you patch.  I started merging this in, and found I do not
> have a "ginstall-info" on my old Red Hat box.  Is there some other
> install that would work?

Mmh... I forgot why I wrote that, then I checked my system and found
that there is an install-info (a Perl script included in the dpkg
package) and a ginstall-info which is part of the texinfo package.

I remember that I had some problem with install-info (debian version)
maybe because it didn't supported some option or some stuff like that
which caused me to waste some time trying to figure out what was wrong
with install-info...

In the case of a redhat system I don't know if there is this problem.

> >I'd also like to add a general install and uninstall rule, but in this
> >case I'd need to know which files have to be installed (and
> >furthermore I'd maybe need to interact with the various Project.ede
> >rather than edit the high level Makefile in the root dir). Let me know
> >if you're interested in such rules.
>   [ ... ]
> 
> You are right, EDE would know what to install.  This is something I've
> never gotten  around to.
> 
> In the grand scheme of things I haven't worked on this because the
> only bits that aren't installed are some of the makefiles and build
> scripts.  (ie, the .el and .elc files are typically installed
> together.)

Hi, thanks again!
-- 
Stefano Sabatini
Linux user number 337176 (see http://counter.li.org)

Thanks for the update.  I'll see what I can figure out.

Eric

>>> Stefano Sabatini <stefano.sabatini-lala@...> seems to think that:
>On date Thursday 2008-04-10 10:31:23 -0400, Eric M. Ludlam wrote:
>> Hi Stefano - 
>> 
>> I apologize for this very late email.  I am finding lots of mail from
>> last summer I did not respond to.  I'm trying to get through all my
>> lost email and found yours.
>> 
>> >>> Stefano Sabatini <stefano.sabatini-lala@...> seems to think that:
>> >Hi Eric and company,
>> >
>> >I'm a great fan of cedet and ECB from more than one year now (and
>> >still trying to understand how many things work... ;-)).
>> >
>> >This patch adds support in the Makefile for an install-info and
>> >uninstall-info rule.
>> 
>> 
>> Thanks for you patch.  I started merging this in, and found I do not
>> have a "ginstall-info" on my old Red Hat box.  Is there some other
>> install that would work?
>
>Mmh... I forgot why I wrote that, then I checked my system and found
>that there is an install-info (a Perl script included in the dpkg
>package) and a ginstall-info which is part of the texinfo package.
>
>I remember that I had some problem with install-info (debian version)
>maybe because it didn't supported some option or some stuff like that
>which caused me to waste some time trying to figure out what was wrong
>with install-info...
>
>In the case of a redhat system I don't know if there is this problem.
>
>> >I'd also like to add a general install and uninstall rule, but in this
>> >case I'd need to know which files have to be installed (and
>> >furthermore I'd maybe need to interact with the various Project.ede
>> >rather than edit the high level Makefile in the root dir). Let me know
>> >if you're interested in such rules.
>>   [ ... ]
>> 
>> You are right, EDE would know what to install.  This is something I've
>> never gotten  around to.
>> 
>> In the grand scheme of things I haven't worked on this because the
>> only bits that aren't installed are some of the makefiles and build
>> scripts.  (ie, the .el and .elc files are typically installed
>> together.)
>
>Hi, thanks again!

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

Thanks for the message.

I checked in two fixes.  One was to just add an autoload cookie to the
missing function, and the other was to not call into the include
decoration mode if it is not loaded.

I haven't had a chance to explicitly test that stuff other than what a
little use I've given it so far.  Hopefully it will work out for you.

I'm don't want to add that require to the minimum features function.
The goal is to make the minimum features as fast as possible.

Thanks
Eric

>>> Dan Kruchinin <dan.kruchinin@...> seems to think that:
>Hi.
>
>I have a problem with cedet from cvs: after my c-mode-hook loads cedet 
>and configures all needful parameters
>(more precisely: it calls semantic-load-enable-minimum-features and 
>semantic-load-enable-code-helpers functions) I get an error
>described in the subject.
>I found out that the problem lies in the macro 
>"defcustom-mode-local-semantic-dependency-system-include-path"
>(semantic/semantic-dep.el, revision #1.9).  After expansion this macro 
>calls mode-local-map-mode-buffers function(common/mode-local.el)
>with 'semantic-decoration-unparsed-include-do-reset(which will be called 
>later from the mode-local-map-file-buffers
>using funcall) as one of its arguments. But function mode-local.el 
>doesn't know anything about 
>semantic-decoration-unparsed-include-do-reset(which actually defined in 
>the semantic/semantic-decorate-include.el), therefore
>'semantic-decoration-unparsed-include-do-reset is seen by 
>mode-local-map-file-buffers as symbol which has an empty symbol-function
>value.
>
>As I can see semantic-decorate-include.el is loaded from only one place 
>- semantic-load-enable-gaudy-code-helpers. But I don't
>want to use this function, because it enables many decorations and 
>functionality that I don't really want to use and see. I only want
>my ecb works well but it doesn't because of described error.
>
>So, may it be reasonable to add (require 'semantic-decorate-include) to 
>the semantic-load-enable-minimum-features?
  [ ... ]

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