Current CEDET version in the trunk

Current CEDET version in the trunk

[Christoph]

Hi Eric and all,

Inspired by the current CEDET discussion, I decided to seriously try out 
Semantic from Emacs 24.0.50 (trunk) today.

Going through the Semantic manual I tried all kinds of things to figure 
out what the current state of the integration with Emacs is.

The Semantic manual is a good way to get started, but there are quite a 
few things that do not work according to the manual and it is not 
obvious how to make it work. For example, using Smart completion (with 
semantic-complete-analyze-inline) the manual says that M-n/M-p select 
the next/previous matches. This does not work. M-n/M-p does nothing. C-h 
k in this mode says, the keys are not assigned to anything.

I assume this is because of the current state of the integration?

Nevertheless, I am liking what I see so far. (External) CEDET always 
seemed like this big monster that was complicated to set up and use, but 
with a tight native integration with Emacs things like Semantic are set 
out to be a killer feature.

Maybe we should better integrate and polish certain features one at a 
time. Semantic seems like a great first choice to me. Semantic-based 
completion and code browsing out of the box. I still have not been able 
to integrate ctags/etags/xyztags satisfactorily and there is always so 
much fiddling involved with tags tables etc. And it's hard to argue that 
IntelliSense is one area where VS and Co. have a competitive advantage 
over Emacs. It just works out of the box.

Is anybody actively working on integrating CEDET better with Emacs right 
now? Or do we just merge the snapshots into the trunk?

Now, going through the manual and playing with it I found quite a few 
inconsistencies/bugs. Where is the right place to report those? The GNU 
bug tracker or the Sourceforge tracker?

Christoph

Re: Current CEDET version in the trunk

[Eric M. Ludlam]

Hi

On 07/13/2010 08:50 PM, Christoph wrote:
> The Semantic manual is a good way to get started, but there are quite a
> few things that do not work according to the manual and it is not
> obvious how to make it work. For example, using Smart completion (with
> semantic-complete-analyze-inline) the manual says that M-n/M-p select
> the next/previous matches. This does not work. M-n/M-p does nothing. C-h
> k in this mode says, the keys are not assigned to anything.

The keymaps used for completion are done via pre-command-hook which 
obscures them from C-h k. I just tried it out and M-n and p work for me 
with CEDET/CVS in Emacs Lisp mode using the idle completion feature.

> Maybe we should better integrate and polish certain features one at a
> time. Semantic seems like a great first choice to me. Semantic-based
> completion and code browsing out of the box. I still have not been able

I tend to agree.  While CEDET/Semantic has some completion UI tools, 
there are others completion UIs that just use the the smart completion 
back-end to calculate values.  (ie - company mode, predictive, hippie 
expand) The Emacs community here has a wide range of choice for 
selecting the 'official' way of doing smart completion inline.  Do not 
assume that the completion engine I wrote is particularly special just 
because it is in the semantic package.  That said, it does have some 
cool features. ;)

> Is anybody actively working on integrating CEDET better with Emacs right
> now? Or do we just merge the snapshots into the trunk?

Lluis is trying to setup an CEDET/bzr repository for me where merges can 
happen more easily.  My most recent papers should be at the FSF now, 
allowing free-flowing patches and merges to occur.

> Now, going through the manual and playing with it I found quite a few
> inconsistencies/bugs. Where is the right place to report those? The GNU
> bug tracker or the Sourceforge tracker?

I use the cedet-devel mailing lists on the CEDET sourceforge site, and 
usually just fix things as they arrive.  If changes go into Emacs, I'll 
eventually get them so that is fine with me also.

Eric

Re: Current CEDET version in the trunk

[Christoph]

Hi Eric,

On 7/13/2010 8:45 PM, Eric M. Ludlam wrote:

> The keymaps used for completion are done via pre-command-hook which
> obscures them from C-h k. I just tried it out and M-n and p work for me
> with CEDET/CVS in Emacs Lisp mode using the idle completion feature.

I just ran M-x semantic-mode after starting Emacs. Nothing else. Do I 
need to configure more than this?

>> Maybe we should better integrate and polish certain features one at a
>> time. Semantic seems like a great first choice to me. Semantic-based
>> completion and code browsing out of the box. I still have not been able
>
> I tend to agree. While CEDET/Semantic has some completion UI tools,
> there are others completion UIs that just use the the smart completion
> back-end to calculate values. (ie - company mode, predictive, hippie
> expand) The Emacs community here has a wide range of choice for
> selecting the 'official' way of doing smart completion inline. Do not
> assume that the completion engine I wrote is particularly special just
> because it is in the semantic package. That said, it does have some cool
> features. ;)

OK, but it worked. I have messed around alot with ctags, exuberant ctags 
(why is there even a difference?), etags (why etags??) and none of them 
seemed to work right always or it was extremely complicated to set up 
and maintain. With Semantic it seemed to work great, right away. It even 
jumped to the definition of a local variable, no problem. Yes, maybe I 
am missing something with the above mentioned tools, but point is, it 
shouldn't be so difficult.

>> Now, going through the manual and playing with it I found quite a few
>> inconsistencies/bugs. Where is the right place to report those? The GNU
>> bug tracker or the Sourceforge tracker?
>
> I use the cedet-devel mailing lists on the CEDET sourceforge site, and
> usually just fix things as they arrive. If changes go into Emacs, I'll
> eventually get them so that is fine with me also.

Whatever you prefer. I can send bug reports to the CEDET mailing list. I 
can also provide patches, but I'd prefer fixing stuff in Emacs then, 
since I already have that source from the Emacs trunk and don't need to 
mess with CVS.

I'll get on the CEDET list and provide an overview of what I have found.

Christoph

Re: Current CEDET version in the trunk

[Lluís]

Eric M Ludlam writes:
>> Is anybody actively working on integrating CEDET better with Emacs right
>> now? Or do we just merge the snapshots into the trunk?

> Lluis is trying to setup an CEDET/bzr repository for me where merges can 
> happen more easily.  My most recent papers should be at the FSF now, 
> allowing free-flowing patches and merges to occur.

Well, I'm still waiting for the people at Sourceforge to fix a directory
permission bug that does not allow me to push into their bzr repository, but
there has been no response so far.

Lluis

-- 
 "And it's much the same thing with knowledge, for whenever you learn
 something new, the whole world becomes that much richer."
 -- The Princess of Pure Reason, as told by Norton Juster in The Phantom
 Tollbooth

Re: Current CEDET version in the trunk

[Eli Zaretskii]

> Date: Tue, 13 Jul 2010 18:50:23 -0600
> From: Christoph <cschol2112@googlemail.com>
> Cc: eric@siege-engine.com
> 
> The Semantic manual is a good way to get started, but there are quite a 
> few things that do not work according to the manual and it is not 
> obvious how to make it work.

Please be sure to submit Emacs bug reports for any discrepancy you
find.

TIA

Re: Current CEDET version in the trunk

[David Kastrup]

Eli Zaretskii <eliz@gnu.org> writes:

>> Date: Tue, 13 Jul 2010 18:50:23 -0600
>> From: Christoph <cschol2112@googlemail.com>
>> Cc: eric@siege-engine.com
>> 
>> The Semantic manual is a good way to get started, but there are quite a 
>> few things that do not work according to the manual and it is not 
>> obvious how to make it work.
>
> Please be sure to submit Emacs bug reports for any discrepancy you
> find.

Given that most discrepancies stem from Emacs not having the full
semantics installed (and this will hopefully change eventually), that
seems like tilting against windmills.  I see little point in changing
the manual to correspond with the current partial semantics.

As long as the plan is to include full semantics in Emacs 24.1.

-- 
David Kastrup

Re: Current CEDET version in the trunk

[Eli Zaretskii]

> From: David Kastrup <dak@gnu.org>
> Date: Wed, 14 Jul 2010 09:06:25 +0200
> 
> > Please be sure to submit Emacs bug reports for any discrepancy you
> > find.
> 
> Given that most discrepancies stem from Emacs not having the full
> semantics installed (and this will hopefully change eventually), that
> seems like tilting against windmills.  I see little point in changing
> the manual to correspond with the current partial semantics.

I didn't say how to fix those bugs, whether by installing more of
Semantic or by changing the manual.  I don't think you or me are
clairvoyant enough to know up front whether such bugs are due to
missing parts of Semantic or to simple omissions and mistakes in the
manual.  The important thing is to report a bug, so it gets fixed
eventually.

Re: Current CEDET version in the trunk

[Chong Yidong]

David Kastrup <dak@gnu.org> writes:

> Given that most discrepancies stem from Emacs not having the full
> semantics installed

Except for the editing of grammar files, most discrepancies are just
bugs, and have nothing to do with not having the full semantics
installed.  They should be reported.

Re: Current CEDET version in the trunk

[Chong Yidong]

Christoph <cschol2112@googlemail.com> writes:

> The Semantic manual is a good way to get started, but there are quite
> a few things that do not work according to the manual and it is not
> obvious how to make it work. For example, using Smart completion (with
> semantic-complete-analyze-inline) the manual says that M-n/M-p select
> the next/previous matches. This does not work. M-n/M-p does
> nothing. C-h k in this mode says, the keys are not assigned to
> anything.
>
> I assume this is because of the current state of the integration?

No, it's a bug.  Please report it as an Emacs bug (since it's not
present in CVS CEDET).

> Now, going through the manual and playing with it I found quite a few
> inconsistencies/bugs. Where is the right place to report those? The
> GNU bug tracker or the Sourceforge tracker?

A large number of these are likely to be leftover bugs from the merge.
Use the GNU tracker.