On 16.04.2020 22:07, Tom Tromey wrote:
>>>>>> ">" == Kamil Rytarowski <n54@gmx.com> writes:
> 
>>> All major BSDs implement PT_GET_PROCESS_STATE, but they differ in
>>> details and want to implement follow-fork functionality differently.
> 

Actually I was wrong in this detail. FreeBSD implements
PT_SET_EVENT_MASK, but not PT_GET_PROCESS_STATE.

>>> gdb/ChangeLog:
> 
>>> 	* inf-ptrace.h (follow_fork, insert_fork_catchpoint)
>>> 	(remove_fork_catchpoint, post_startup_inferior)
>>> 	(post_attach): Move...
> 
> I guess these aren't used on Linux, but what about rs6000-nat.c?  Those
> are the only non-BSD users of inf_ptrace_target that I can see.
> 
> I'm not a big fan of the current code, because makes methods conditional
> on this define.  It would be better I guess to have an intermediate
> subclass for this situation.  That would make it clear what code does or
> does not rely on this.
> 

Personally, I would like to see more separation of OS specific features
into OS-private files.

> You don't have to do that, but I don't think we can put this in without
> knowing about the rs6000-nat case.
> 
> Tom
> 

IBM RS/6000 is AIX and it does not support PT_GET_PROCESS_STATE. At
least not in the versions I could check.

There was however support for PT_GET_PROCESS_STATE in HPUX, but HPUX
handling was removed from GDB in 2015.