This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

RE: Adding platform extensions?


Debugging info?
hmm...that would work, to some extent....I don't want the final executable
to have debugging info taked in
too bloated....

Here is how my binaries are layed out:
---------
Image Header  (Currently PE, looking at ELF)
---------
Object Table
---------
Interface Table
---------
Compiled Code

currently to come up with that it is very redundant and there are some
pitfalls to a clumbersom programming style like that.  BUT, the
object/interface technologies in the kernel work great so I don't plan on
changing them to adapt a different approach....I need compiler tools
modified so that they work with the original idea.

Now, back to my original question back a few days ago.
If I were to go about adding support to the compiler, how would I do this?
I see table upon table of numbers in the parsing engine....this code appears
alien in nature, anyone want to give me hints at what I would try?

Of course, if anyone wants in on my OS project and wants to help organize a
small porting group for making the tools, I'd obviously give all the
information I need to assist and try to help out too, but I am not expecting
this.

-Dan

-----Original Message-----
From: Jamie Lokier [mailto:jamie.lokier@cern.ch]
Sent: Tuesday, November 16, 1999 9:43 AM
To: Dan Guisinger
Cc: 'Mike Stump'; gcc@gcc.gnu.org
Subject: Re: Adding platform extensions?


Dan Guisinger wrote:
> One of the reason I don't think that C++ can do it on its own is that the
> structures store more than just a Vtbl, they have several SwCom specific
> data structures in front of it which are hidden to the programmer but
exist
> to track the object, all interface instances, and other data that is
related
> to upkeep of the SwCom subsys and integrety of the system.

I have a similar problem trying to get decent garbage collection...

> So, how about this:
> Is there a way where I can pull the size of a structure from a OBJ file,
if
> the structure is not created statically?
> That is, if the structure is atleast defined, does its size get stored
into
> the OBJ file so it can be externally referenced?

All the information should be available as debugging info.  Just compile
with debugging on.

-- Jamie


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]