On Sun, Jan 31, 2010 at 1:52 PM, John David Anglin <dave@xxxxxxxxxxxxxxxxxx> wrote: >> Could you explain the change in more detail? > > The change is to set fs->signal_frame = 1 in signal frames in > pa32_fallback_frame_state. This affects the determination as to whether > the IP is before or after the first not yet fully executed instruction > in a frame. This feature was added in 2006 by Jakub Jelinek, but > pa32_fallback_frame_state was not changed. OK. > This fixes unwinding in Ada (corner case). It probably fixes corner > causes for other languages. However, it breaks a number of Java testcases. > The problem is compensated for in the definition used for MAKE_THROW_FRAME > in libjava/include/pa-signal.h and hppa-signal.h. > > The Ada testcase that exposed this problem is new for 4.5. > > I'd hoped to bump the libgcc_s version so as to not break existing > applications, but that appears impossible. No, unfortunately the implementation of pthread_cancel() needs to use the unwinder, and that means glibc assumes that the latest libgcc_s.so can always unwind old code using some sort of fallback. Cheers, Carlos. -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
