[RFC] GDB patches for hw watchpoints - revised

Wu Zhou woodzltc@cn.ibm.com
Wed Feb 8 05:35:00 GMT 2006


Hi Daniel,

Sorry for the delayed reply.  I am just back from a vacation.

On Wed, 1 Feb 2006, Daniel Jacobowitz wrote:

> On Wed, Jan 25, 2006 at 04:34:14PM +0800, Wu Zhou wrote:
> > I am also thinking of replace the macro 
> > TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT (SIZE) with 
> > TARGET_REGION_OK_FOR_HW_WATCHPOINT(ADDR, SIZE).  Thus the code will seem 
> > to be more clean: we will only have one macro to see if the target region 
> > is ok for watchpoint monitoring.  Following this way, function 
> > default_region_ok_for_hw_watchpoint will also return back to its original 
> > implementation.  What is your thought on this?
> 
> I'm not sure what you mean.  After these patches, the only reference to
> TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT is in gdbint.texinfo (which I'd
> appreciate if you fixed, in a separate patch - thanks in advance).
> What are the two ways now?

What I mean is the second patch below, which you had said ok. :-) 
That is the only way I can thought of at that time.  Now that you had said 
ok, I don't need to find a second way.  :-)

BTW.  I will fix the reference to TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT 
in gdbint.texinfo in a separate patch after commiting this.

> 
> On Wed, Jan 25, 2006 at 11:18:49AM +0800, Wu Zhou wrote:
> > OK to commit?
> > 
> > 2006-01-22  Ben Elliston  <bje@au1.ibm.com>
> > 	    Wu Zhou  <woodzltc@cn.ibm.com>
> > 
> > 	* ppc-linux-nat.c (PTRACE_GET_DEBUGREG, PTRACE_SET_DEBUGREG,
> > 	PTRACE_GETSIGINFO): Define.
> > 	(last_stopped_data_address): New.
> > 	(ppc_linux_check_watch_resources): New function.
> > 	(ppc_linux_region_ok_for_hw_watchpoint): New function.
> > 	(ppc_linux_insert_watchpoint): New function.
> > 	(ppc_linux_remove_watchpoint): New function.
> > 	(ppc_linux_stopped_data_address): New function.
> > 	(ppc_linux_stopped_by_watchpoint): New function.
> > 	(_initialize_ppc_linux_nat): Set the above hardware watchpoint
> > 	related target vectors.
> > 	* rs6000-tdep.c (rs6000_gdbarch_init): Set PPC architectures
> > 	to have nonsteppable watchpoint.
> > 	* target.c (default_region_ok_for_hw_watchpoint,
> > 	debug_to_region_ok_for_hw_watchpoint): New prototypes.
> > 	(update_current_target): Inherit to_region_ok_for_hw_watchpoint
> > 	and set default to_region_ok_for_hw_watchpoint.
> > 	(default_region_ok_for_hw_watchpoint): New function.
> > 	(debug_to_region_ok_for_hw_watchpoint): New function.
> > 	(setup_target_debug): Set to_region_ok_for_hw_watchpoint of 
> > 	debug_target.
> > 	* target.h (struct target_ops): Add a new target vector 
> > 	to_region_ok_for_hw_watchpoint.
> > 	(TARGET_REGION_OK_FOR_HW_WATCHPOINT): Define this if it is not
> > 	defined anyplace else.
> 
> On Wed, Jan 25, 2006 at 04:34:14PM +0800, Wu Zhou wrote:
> > 2006-01-25  Wu Zhou  <woodzltc@cn.ibm.com>
> > 
> > 	* breakpoint.c (TARGET_REGION_OK_FOR_HW_WATCHPOINT): Delete.
> > 	* config/i386/nm-i386sol2.h (TARGET_REGION_OK_FOR_HW_WATCHPOINT): New.
> > 	(TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT): Delete.
> > 	* config/mips/nm-irix5.h (TARGET_REGION_OK_FOR_HW_WATCHPOINT): New.
> > 	(TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT): Delete.
> > 	* config/sparc/nm-sol2.h (TARGET_REGION_OK_FOR_HW_WATCHPOINT): New.
> > 	(TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT): Delete.
> > 	* inf-ttrace.c (inf_ttrace_region_ok_for_hw_watchpoint): New.
> > 	(inf_ttrace_region_size_ok_for_hw_watchpoint): Delete.
> > 	(inf_ttrace_target): Delete to_region_size_ok_for_hw_watchpoint and
> > 	add to_region_ok_for_hw_watchpoint.
> > 	* s390-nat.c (s390_region_size_ok_for_hw_watchpoint): Delete.
> > 	(s390_region_ok_for_hw_watchpoint): New.
> > 	(_initialize_s390_nat): Delete to_region_size_ok_for_hw_watchpoint
> > 	and add to_region_ok_for_hw_watchpoint.
> > 	* target.c (default_region_size_ok_for_hw_watchpoint, 
> > 	debug_to_region_size_ok_for_hw_watchpoint): Delete prototype.
> > 	(update_current_target): Delete to_region_size_ok_for_hw_watchpoint
> > 	inheritance and default_region_size_ok_for_hw_watchpoint.
> > 	(default_region_ok_for_hw_watchpoint): If len is less than or equal
> > 	the length of void pointer, return ok.
> > 	(default_region_size_ok_for_hw_watchpoint): Delete.
> > 	(debug_to_region_size_ok_for_hw_watchpoint): Delete.
> > 	(setup_target_debug): Delete to_region_size_ok_for_hw_watchpoint.
> > 	* target.h (struct target_ops): Delete 
> > 	to_region_size_ok_for_hw_watchpoint.
> > 	(TARGET_REGION_SIZE_OK_FOR_HW_WATCHPOINT): Delete.
> 
> These patches are both OK.  You might want to combine them - it's a
> much smaller diff :-)  But it doesn't matter since you've already got
> them separated out.

Thanks for reviewing that.  I would like to commit them one by one, thus I 
don't need to think about how to re-describe them.  :-)

I also think that it is clearer to differentiate the purpose of these two 
patches.  Combining them together seems a little confusing to me.

Best Regards
- Wu Zhou



More information about the Gdb-patches mailing list