On 07/14/2010 09:31 PM, Óscar Fuentes wrote:
> I tested CEDET for about two hours on a small (30K LOC) C++ project.
>
> As an Elisp package that tries to make sense of such a complex language
> as C++ is, the things it can do so far are impressive. Congratulations.
Thanks
> The version used is what comes with Emacs' development sources. I've
> found quite a few glitches, bugs and all other types of misbehavior. One
> that surprised me was that, as far as I know, there is no procedure for
> showing a list of function/method overloads with
> semantic-ia-show-summary, only the first candidate is shown in the
> minibuffer. Seems that constructors are not well supported by Semantic,
> too.
This came up recently, and a new function was added called
'semantic-ia-show-variants' which works pretty well. Almost all the
`semantic-ia-*' functions are prototypes of perhaps one key
infrastructure piece. I think a 'show me everything there is to know'
or 'show me anything interesting' type function is needed to merge
several of these features together. That is a matter of design at this
point.
> Another blocking issue is that Semantic gets easily confused by a
> sequence of chained method calls:
>
> x.foo().bar()
>
> with the point on `bar', Semantic acts as if the method calls didn't
> exist and insists on saying that only `x' is there.
In a long chain, if Semantic fails on any one item, it stops where it
gets lots. In this case, all it means is that 'foo' wasn't found. You
can use semantic-analyze-debug-assist to see what might be up. This is
also helpful in reporting bugs in case everything is discoverable, but
there is instead an analyzer bug.
> Finally, do you recommend to ditch the CEDET that comes with Emacs and
> switch to the development version from your CVS repo? It would be great
> to use CEDET on my daily work and report bugs based on current
> development state, not on what it was half a year ago. Are some of the
> issues mentioned above fixed on the CVS sources? Do you prefer bugs
> reports here or on SourceForge's bug tracker?
I don't know what is an Emacs integration bug, and what is something I
may have fixed. I can only really help with CEDET/CVS reports. Chong,
however, needs reports on Emacs's integrated version of CEDET.
I would say that if you are using some new features (ie - show variants)
or find bugs deep in the analyzer, CEDET/CVS is the way to go since I
have fixed many since the integration. If there are more basic issues,
the Emacs tracker is probably needed.
Eric
|