On 09/06/2017 11:11 PM, Samuel Thibault wrote:
> Pedro Alves, on lun. 04 sept. 2017 13:14:33 +0100, wrote:
>> On 08/27/2017 07:41 PM, Samuel Thibault wrote:
>>> Since hurd's baf7e5c ('hurd: Use polymorphic port types to return some
>>> rights.'), some RPCs prototypes have changed, gdb needs the
>>> corresponding update.
>>>
>>> * gdb/gnu-nat.c (S_proc_getmsgport_reply, S_proc_task2proc_reply,
>>> S_proc_pid2proc_reply): Add `mach_msg_type_name_t type' parameter.
>>
>> Say someone downloads some prebuilt Debian Hurd image or some
>> such and wants to build newer gdb on that system. I assume that that would
>> be
>> broken with this change?
>
> Yes.
>
>> What's the policy regarding building ToT gdb on non-ToT Hurd systems?
>> Is the intention to only ever support building ToT gdb with ToT Hurd?
>
> ATM we don't really support backward compatibility for mixtures of
> versions.
OK, seems to me that raises bar of entry to gdb/hurd hacking,
but really up to you guys. A few years back, when I did some
across-all-gdb-targets changes, I used the prebuilt Debian
Hurd qemu image to do&test the corresponding Hurd changes. Guess
I may have been lucky then to not hit some version skew.
Thanks,
Pedro Alves
