Hello,

There was a presentation about ARM in my university and I asked them about the errors when accessing unaligned memory.

The good news is that they said that the latest generation of ARM processors supports unaligned memory access.

They also asked me to send them some code which illustrates the problem, which they claim should not exist because the compiler should make sure everything is aligned.

So, does anyone have a particularly illustrative example?

thanks,

They also asked me to send them some code which illustrates the problem, which they claim should not exist because the compiler should make sure everything is aligned.

FPC does too as long as you don't force the compiler to use a certain memory layout like using the packed keyword or casting unaligned pointers like plongint(100003)^:=1234;

Actually, FPC handles even packed types correctly by assuming that they are unaligned so the example below is the worst case but there the compiler can't do anything but the user can apply the unaligned keyword to tell the compiler that the data is unaligned:

Unaligned(plongint(100003)^):=1234;

or casting unaligned pointers like plongint(100003)^:=1234;

Indeed, but did anybody investigate why this goes wrong?

what do you think about http://bugs.freepascal.org/view.php?id=12137 ? IMO the unaligned directive should not be required in this case to "fix" the bug.

Regards, Bernd.