[eresi-dev] Is etrace able to trace (GCC)c++ simbols ?

[Julien Vanegue]

Hi Marius

The problems I see is if your target binary has been so optimized that there
is nothing to trace anymore (not counting the fact that we currently only
trace functions with arguments, which is a limitation that is really easy to
lift as most of the problem in our tracing technique resides in argument
counting, whose code is located in libelfsh/ia32.c search for "argcount",
and in the case of a function without parameter, we return 0 (which is
interpreted as an error by the calling function, so we safetly discard this
function for tracing)

Btw, Can you reproduce the test in testsuite/tracing/ ? (this is known to
work)

If you want to contribute a symbol mangling function to make etrace support
C++ binaries, I could tell
you easily how to integrate it into libelfsh :) I can remove the 0 args
limit if you have an interest.

Let me know

Julien



On Tue, Sep 22, 2009 at 12:01 PM, Groleo Marius <groleo at gmail.com> wrote:

> I looked around testsuite/tracing and found host.c
> I player with it and got it to this:
>
>
> #include <stdio.h>
>
> void  testcrypt()
> {
> }
>
> int main()
> {
>   char path[256]={};
>   putchar(path[0]);      /*  Needed for successful `traces add`  */
>   return 0;
> }
>
>
> It seems that I have to have a stack space allocated and used as parameter.
> Don't know if it makes sense... :D
>
>
>
> On Tue, Sep 22, 2009 at 12:43 PM, Groleo Marius <groleo at gmail.com> wrote:
>
>> Seems like I can't paste messages bigger than 40k, so:
>> http://pastebin.ca/1575047
>>
>>
>>
>>
> --
> Regards, Groleo!
>
> _______________________________________________
> eresi-dev mailing list
> eresi-dev at lists.einherjar.de
> http://lists.einherjar.de/mailman/listinfo/eresi-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.einherjar.de/pipermail/eresi-dev/attachments/20090922/130471cc/attachment.html>