Re: [cedet-semantic] Completion-UI support for Semantic

[Toby C. <ts...@ca...>]

Hmmmm... M-TAB is bound in `completion-map', which is added to the
`minor-mode-map-alist' when the completion-ui library is loaded. So
either something is overriding that (a minor-mode map that appears
earlier in the alist?), or the completion-UI library isn't actually
being loaded at all.

Easy test: try "M-x complete-word-at-point". If that fails,
completion-UI isn't loaded. If it works, try "M-/" (or better still,
"C-h k M-/" and check whether it's bound to `complete-word-at-point').
If that works, the M-TAB binding is being overridden by something else.
Use "C-h k M-TAB" to see what might be doing this.

In other words, I still believe this to be a simple configuration issue,
and would like to rule this out before expending further effort on it :)

I don't usually like adding `require' statements within function bodies,
but perhaps in the case of the EmacsWiki code snippets, it might be a
good idea. It would at least give an obvious error message and rule out
some trivial issues.

Toby


Eric M. Ludlam wrote:
> I just noticed that the default binding of M-TAB is `complete-symbol'
> which has doc that says:
> 
> ---
>  Perform tags completion on the text around point.
>  Completes to the set of names listed in the current tags table.
> ---
> 
> so it sounds like the keybinding M-TAB was never rebound to
> completion-UI.
> 
> Eric
> 
>>>> Toby Cubitt <ts...@ca...> seems to think that:
>> Hi Eric,
>>
>> Thanks for finally injecting some intelligent input into my wild
>> speculations :)
>>
>> Completion-UI doesn't default to any kind of completion. Until you set
>> the `completion-function' variable, it does absolutely nothing at all.
>> So somewhere, someone would have to explicitly set `completion-function'
>> to the etags wrapper function (in a .emacs?). But running
>> `completion-setup-semantic' sets `completion-function' (buffer-locally),
>> overriding whatever was there before. So I don't see how it
>> completion-UI could be calling the etags wrapper instead of the Semantic
>> one, unless someone told it to.
>>
>> I have this vague recollection that Semantic might use etags in some
>> major-modes (elisp?). Does that recollection have any reflection in
>> reality? Could that explain why it's complaining?
>>
>> Otherwise, the answer lies in a backtrace...
>>
>> Toby
>>
>>
>> Eric M. Ludlam wrote:
>>> Hi,
>>>
>>> The message:
>>>
>>> No tags table loaded; Try M-x visit-tag-table"
>>>
>>> is in complete-tag in etags.el.
>>>
>>> I assumed it was that the semantic setup for completion ui was not
>>> reconfigured as expected, and completion-ui was defaulting to etags
>>> completion instead.
>>>
>>> Eric 
>>>
>>>>>> Toby Cubitt <ts...@ca...> seems to think that:
>>>> Hi,
>>>>
>>>> There's little more I can add. "No tags table loaded" is not a
>>>> completion-UI error message (grep the code if you don't believe me), so
>>>> something else is responsible. And I can't reproduce this problem. I
>>>> still suspect there's something not quite configured right in your buffer.
>>>>
>>>> If the code is throwing a lisp error (rather than just printing a
>>>> message), you could `toggle-debug-on-error', repeat the steps that
>>>> produce the error, and look at the backtrace to show you the function
>>>> calls that caused the error. I still expect that it's one of the
>>>> Semantic functions, but this would confirm it.
>>>>
>>>> Otherwise, you could edebug the completion-UI Semantic support code, and
>>>> step through it to figure out where the problem occurs.
>>>>
>>>> If you still can't figure out what's going on, you could send me (off
>>>> list) the contents of the backtrace, or details of the results of
>>>> stepping through the code.
>>>>
>>>>
>>>>> I'm sure semantic worked for me, as i can use company-mode properly and
>>>>> semantic-ia-complete-symbol-menu to pop up a menu of completions.
>>>> Company mode's Semantic support code is very, very similar to the
>>>> completion-UI support code...except that it's surrounded by
>>>> "(ignore-errors...)". So the fact that company-mode works doesn't tell
>>>> me much :)
>>>>
>>>> The fact that `semantic-ia-complete-symbol-menu' works does seem to
>>>> indicate that Semantic is correctly enabled, but I'm not familiar enough
>>>> with Semantic to know what's going on without being able to reproduce
>>>> the error myself...
>>>>
>>>>
>>>>> what's more, i think allowing users to re-bind the function of M-TAB with
>>>>> other keys will be more flexible since some users may use M-TAB to switch
>>>>> between windows.
>>>> Errmmmm...did you look at completion-UI? I know it's not documented very
>>>> well, but it does most things in the conventional Emacs way...including
>>>> binding keys in keymaps (imaginatively called `completion-map' etc.)
>>>> which can be rebound to your heart's content.
>>>>
>>>> Moreover, M-/ is bound to the same command as M-TAB, for precisely the
>>>> reason you cite.
>>>>
>>>> Toby
>>>>
>>>>
>>>>
>>>>
>>>>>> Hi,
>>>>>>
>>>>>> bread wrote:
>>>>>>> I tried M-TAB after M-x completion-setup-semantic , but i was told that
>>>>>> "No
>>>>>>> tags table loaded; Try M-x visit-tag-table"$B!#(B
>>>  >>> This is a CEDET error message, so has nothing to do with completion-UI.
>>>>>> It sounds like you haven't enabled Semantic in the buffer. I'm not a
>>>>>> Semantic expert, but you'll probably want to enable
>>>>>> `semantic-idle-scheduler-mode' at least.
>>>>>>
>>>>>> If it still doesn't work, try getting Semantic to work in the buffer
>>>>>> first, before trying to enable completion-UI. I recommend reading the
>>>>>> INSTALL file that accompanies CEDET carefully...
>>>>>>
>>>>>> HTH,
>>>>>>
>>>>>> Toby