On 03/16/2011 11:06 AM, Tobias Gerdin wrote:
> Den 2011-02-28 05:05:51 skrev Eric M. Ludlam<eric@...>:
>> One of my new hobbies has been writing apps for my android device. As
>> such, I've now posted a new "android" branch in bzr that adds a pile of
>> support for writing code with the android SDK. Some of this has been
>> hanging in my sandbox for a long time, so I'm glad to get this branch
> This seems pretty neat. I didn't get it to work when I tried it (may post
> more info on this if I decide to play with it more).
I posted a few small patches today from the last couple weeks. Perhaps
one of those will help, though I don't know what issue you ran into.
> I was wondering if it wouldn't be possible to use this strategy for C but
> by extracting the relevant information from debug info? I'm not very
> familiar with debug info formats, but some quick googling of DWARF does
> seem to indicate that it has what I assume would be needed (return type,
> function name, parameters and their types).
Yes, doing so isn't a big deal. Actually, it took me the better part of
a day to figure it out, but at least it is a paved road now.
Anyway, javap outputs real java code, and I just ran the regular java
parser. If you do this for C/C++, it will need need to do the same, or
you will need some new parser to extract the contents. I experimented
w/ a live GDB session several years ago, and couldn't figure out a
reliable way to ask GDB questions, so if there is something better, that
would be cool. If someone hacked up a script to transform the output of
something like nm into useful to Emacs stuff, then the rest is doable.
This would be useful for C/C++ in that most Semantic stuff targets
headers, which annoys people who want to jump to the src instead.
Having this as a quick way to target some src file would be nice, and
probably wouldn't need the type of features in the javap stuff.