>>> Marcus Harnisch <marcus.harnisch@...> seems to think that:
>After realizing that a fully functional parser would be extremely hard
>to create, I decided to stick with what I have for now. Works fine so
>Now there are some remaining issues which I would like to bring up.
>1. I have two structs (classes) defined in my test file. Both of them
> have a field named `field1'. The types of the respective fields in
> each class are different. eldoc keeps displaying the type of the
> first definition, no matter whether the cursor is in the other
This is a long standing problem with the eldoc implementation. It
shows the first occurrence it finds.
The CVS version of semantic has a new auto parse mode, and a new
idle-summary mode. This uses the semantic analyzer to correctly find
the "field1" that is appropriate.
> struct or not. Also, as soon as the timer function gets called, the
> machine gets really busy until I finally hit C-g which gives me the
> message `itimer "itimer-0" quit'. Backtraces differ each time I
I'm not sure about that one. Searching very large data sets can be
pretty slow, but that doesn't sound likely in your case.
> create one to get more information. It doesn't seem to be my own
> funny code. I have the same problem when I just edit any
> semantic-supported file. Could be eldoc itself. The XEmacs version
> doesn't seem to be quite up-to-date.
>2. I was trying to use Speedbar analysis and get only `Empty
> Display' although I am sure I did get something more useful at
> some earlier stage. Just don't remember how.
> At what stage of development are the functions in the Analysis
The analysis tool is pretty good at C code when your entire project
is available. Getting it to work for new languages requires some
extra footwork, including making sure all overload functions in
semantic-ctxt.el have proper implementations, and making sure that
all tags created follow a pretty strict set of defined guidelines for
An easy way to try it out is to just use M-x
semantic-analyze-current-context, and see what it says.
>3. Macros and templates. Anyone done that, yet? In my language, quite
> in contrary to C/C++, the replacement text of a macro must be valid
> code. That should make things easier. I was wondering if one could
> keep a table with macro names and their replacment texts and have
> the lexer temporarily splice the text into the stream. Just to
> generate the token stream for the parser.
> Templates are far more difficult to deal with in the language I am
> working with, but also not as common in casual use.
Macros require a pre-processor which is not available in semantic.
Perhaps someday it will be. We take short cuts in the lexical
analyzer that is efficient in space, but prevent an easy
implementation of a preprocessor.
Eric Ludlam: zappo@..., eric@...
Home: http://www.ludlam.net Siege: http://www.siege-engine.com
Emacs: http://cedet.sourceforge.net GNU: http://www.gnu.org