>In which case the most likely cause is a compiler bug.
>(I've had occasional problems at different with what seems to be multiple use
>of the same register, evident either as things like ldr r3, [r3],#3 or
>once with loading a short int using 2 lrdb instructions both targeted at
>the same register, all when building perl.
I'm reasonably confident that both of those problems are fixed now. The
former would certainly elicit a warning from gas, and I don't see any of those
in the log. But there could well be other, similar bugs still lurking.
>Is it possible to arrange for the buildd either to use a different
>optimisation level, or to temporarily use a new (or older) compiler?
I'd rather not tinker with the buildd's environment, but I'm happy to try
compiling with different sets of options.