Re: Get rid of ATTACH_NO_WAIT

[Pedro Alves <pedro at codesourcery dot com>]

Hi Joel,

A Thursday 03 July 2008 17:47:44, Joel Brobecker wrote:
> Hi Pedro,
>
> > 2008-06-28  Pedro Alves  <pedro@codesourcery.com>
> >
> > 	* config/i386/nm-cygwin.h (ATTACH_NO_WAIT): Delete.
> > 	* config/i386/nm-i386gnu.h (ATTACH_NO_WAIT): Delete.
> >
> > 	* target.h (struct target_ops): Add to_attach_no_wait member.
> > 	(target_attach_no_wait): New.
> > 	* target.c (update_current_target): Inherit to_attach_no_wait.
> >
> > 	* infcmd.c: Replace ATTACH_NO_WAIT compile time check by
> > 	target_attach_no_wait runtime check.
> >
> > 	* gnu-nat.c (init_gnu_ops): Set to_attach_no_wait in gnu_ops.
> > 	* win32-nat.c (init_win32_ops): Set to_attach_no_wait in
> > 	win32_ops.
>
> I will be glad to see this macro go, and overall the patch looks good
> to me.
>
> One thing I had to think about a little was whether this property
> should be inherited or not (see target.c:update_current_target()).
> I'm still not sure, but I think it should.  

I think it should.  Normally, I expect the debug API to attach to a
process to be implemented by the process_stratum target, even if
the thread_stratum is pushed already for whatever reason, and needs
to do poke the just attached process.

> Imagine that we had 
> a thread stratum on win32. Wouldn't you lose the attach_no_wait property
> when this thread stratum got pushed on the target?  Until we find a
> target where the process and thread strata need a different setting,
> I think it's safer for now to make it inheritable. What do you think?

Hmm, I'm already doing it?  Or did I miss anything?

-- 
Pedro Alves