[cfe-dev] absolute locations of functions?

[ret val]

The end goal for this(and another post to this list) is to general a file
that contains all functions and the function calls they make, along with the
addresses(of each) in the binary. Having names would be great for debugging.
I figured at the very least I can get the function calls as a plugin.

Originally I was trying to modify the dot-callgraph pass, but modying the
GraphWriter to be more generic than a dot file was a bit tricky. As a plugin
seems more convent also.

On Thu, Jun 9, 2011 at 2:07 PM, David Chisnall <csdavec at swan.ac.uk> wrote:

> Clang doesn't have this information.  I think you can use the Object stuff
> in LLVM to get it after the binary has been generated, but it's not trivial,
> because this doesn't happen until after the linker has run, and in
> nontrivial cases the linker is not run by the same program invocation as the
> compiler.
>
> What problem are you actually trying to solve?
>
> David
>
> On 9 Jun 2011, at 19:04, ret val wrote:
>
> > Is there a direct way to get the location in the binary, as a plugin? I
> think that would be a nicer form for me, if possible.
> >
> > Thanks
> >
> > On Tue, Jun 7, 2011 at 1:41 PM, David Chisnall <csdavec at swan.ac.uk>
> wrote:
> > On 7 Jun 2011, at 18:11, ret val wrote:
> >
> > > Is there anyway to write a pass/plugin so I can print the locations of
> all functions? Can anything run after the linker? I'm not sure what I should
> be reading up on to tackle this.
> >
> > If you mean their location in the resulting binary, then you don't need a
> plugin to do this: the nm utility will do it for you.  It will print the
> location of all symbols, but you can use the example clang plugin to print
> the names of all functions and then use this to filter the output of nm.
> >
> > If you mean that location in the address space of a running program, then
> this is not possible with a clang plugin, because this location is not
> defined until after the program / library has been loaded, and on modern
> systems will be randomised on each program run as a security precaution.
> >
> > David
> >
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20110609/399cdefb/attachment.html>