Okay, I just had an LED moment (light bulbs are so passe' now...)!
What if you could use the other Intel Core 2 Duo processor to
mirror/emulate the other when debugging a program?
Clearly, both processors are the same, right? So what if you could
use the 2nd processor to 'pre-step' into a line of code and see if it
blows up BEFORE you execute the same line of code on the 1st
processor? For example, take the following C bit of code...
#include <stdio.h>
int main(int argc, char *argv[])
{
int *ptr, myArray[0, 1, 3, 2];
init(); // some sort of setup
ptr = myArray;
// deliberate bug here...
ptr++;
while (*ptr)
{
printf("*ptr = %d\n", *ptr);
printf("ptr = %x\n", ptr);
ptr++;
}
}
While we know this will blow up (it'll step off the end of
'myArray'), the computer doesn't... until its too late. But let's
step into this using our fictitious, new version of gdb. We step up
to the first "ptr++;" line. So far, so good. Now we single-step over
it. Both processors execute the instruction, but - while the first
stops, the second one goes ahead and pre-executes the next
instruction for us (while (*ptr), to be exact.) Again, so far, so good.
We now step through the loop a few times. Once the 2nd "ptr++;"
line executes the third time, gdb will now present us with a warning
before stepping back to the while statement:
**gdb: EXC_NO_ACCESS will occur on next statement unless *ptr
modified to point into 'myArray'.**
Now we can see what the machine (processor) state is BEFORE it
blows up, not just after (when we get the "Your Application has
crashed..." dialog, and its core dump.) Of course, this could be
extended to pre-execute any number of machine instructions (those
making up a line of RB code, for example...), stopping as before on
the statement prior to the problem.
Another advantage: since the processor mirrors the other, we can
dump its locals/call stack/etc... without modifying the one in the
main processor (in other words, we can measure without disturbing
what we're measuring.)
Anyway, just dreaming... carry on.
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives:
<http://support.realsoftware.com/listarchives/lists.html>
