This is the mail archive of the gcc-help@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]
Other format: [Raw text]

Re: L4Ka. Regression with 4.7 and 4.8 ?


On Wed, Jun 19, 2013 at 5:59 AM, BERTRAND Joël
<joel.bertrand@systella.fr> wrote:
>
> If I rebuild the same kernel with the same options, it just runs as
> expected. I have no idea to investigate.

Unfortunately, neither do we.  You know more about the kernel than we
do.  Debugging failed-boot problems is indeed among the hardest of
compiler problems to track down.

This could of course be a compiler bug, but the more common cause is
that improved compiler optimization has found some bug in the kernel
code where the kernel relies on undefined behaviour.  Or, something
has changed in the way that asm statements work.  Or, the kernel is
relying on specific inlining semantics that have changed.

Some steps you can try are: compile the kernel without optimization.
Compile the kernel with optimization but with the -fno-strict-aliasing
option.  Boot the kernel in a VM where you can trace the instruction
sequence; compare the sequence for a working and non-working kernel
and look for where it goes wrong.

Ian


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