Re: [PATCH v7 0/8] Support software single step and conditional breakpoints on ARM in GDBServer.

[Antoine Tremblay <antoine.tremblay@ericsson.com>]

On 12/10/2015 10:59 AM, Yao Qi wrote:
> Antoine Tremblay <antoine.tremblay@ericsson.com> writes:
>
>> In this v7:
>>
>> * New patch 3 changes magic numbers for syscall check to numbers computed
>> from xml-syscall.
>>
>> * Patch 4 now uses the new xml-syscall function to compare syscalls.
>> It also uses the INT_REGISTER_SIZE and ARM_PC_REGNUM defines.
>>
>> * Patch 5: Fixed Makefile changelog.
>>
>> * Patch 6:
>> ** Cleanup where arm-get-next-pcs.h is included.
>> ** Move unrelated functions and macros from arm-get-next-pcs.h/.c to arm.h/.c.
>> ** Fixed arm/thumb_deal_with_atomic_sequence calls.
>> ** Fixed changelog Makefile entries.
>> ** Fixed changelog configure.tgt/srv entries.
>> ** Use __NR_sigreturn and __NR_rt_sigreturn to check for sigreturn syscall
>> in GDBServer.
>
> Hi Antoine, thanks for the update...
>
> FYI, I am reviewing your patches and trying to run various regression
> tests for them.  Even without your patch series, there is a regression
> (in gdb.base/advance.exp) in aarch64 mutli-arch debugging that is,
> breakpoint on main isn't hit when aarch64 GDBserver debugs arm program
> in arm mode (everything works fine if the arm program is in thumb mode).
> I suspect that it is related to my recent patch turning Z0 packet on for
> aarch64 mutli-arch debugging.  I really want to clean the room first
> before we add news things in, so I'll fix the regression first, and then
> come back to review your patches.
>

Sure, thanks for letting me know.

Regards,
Antoine