Am Mittwoch, den 02.10.2013, 22:04 +0200 schrieb Jan Lübbe: > On Wed, 2013-10-02 at 21:45 +0200, Lucas Stach wrote: > > Am Mittwoch, den 02.10.2013, 21:30 +0200 schrieb Jan Luebbe: > > > The function am33xx_get_cpu_rev may be called before barebox_arm_entry(), > > > so we need to avoid switch statements. > > > > Uhm, could you please be more verbose on _why_ we need to avoid switch > > statements at this stage? I might be running into the same failure with > > Tegra at some point if there's a real issue. > > I believe the problem is that the table is accessed using absolute > addresses. Before the call to barebox_arm_entry() barebox may be running > at a different address. > Ah, so it's producing a jump table with absolute offsets even for this rather trivial switch statement? I can see how this would cause issues. How about using "-fno-jump-tables" for the objects that are referenced in the early startup code? This may allow you to drop this code workaround. Regards, Lucas _______________________________________________ barebox mailing list barebox@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/barebox
