From mboxrd@z Thu Jan  1 00:00:00 1970
From: Mark Kettenis <kettenis@chello.nl>
To: jimb@redhat.com
Cc: gdb-patches@sources.redhat.com
Subject: Re: RFA: allocate regsets on gdbarch obstack
Date: Thu, 20 May 2004 12:32:00 -0000
Message-id: <200405201231.i4KCVx0P070862@elgar.kettenis.dyndns.org>
References: <vt2zn83y4fs.fsf@zenia.home>
X-SW-Source: 2004-05/msg00604.html

   From: Jim Blandy <jimb@redhat.com>
   Date: 20 May 2004 01:53:59 -0500

   Index: gdb/regset.h
   ===================================================================
   RCS file: /cvs/src/src/gdb/regset.h,v
   retrieving revision 1.2
   diff -c -p -r1.2 regset.h
   *** gdb/regset.h	20 May 2004 00:53:06 -0000	1.2
   --- gdb/regset.h	20 May 2004 05:19:23 -0000
   *************** struct regset
   *** 51,60 ****
	function is COLLECT_REGSET.  If the regset has no collect function,
	pass NULL for COLLECT_REGSET.

   !    The object returned is allocated using xmalloc.  */
   ! extern struct regset *regset_xmalloc (void *descr,
   !                                       supply_regset_ftype *supply_regset,
   !                                       collect_regset_ftype *collect_regset);


     #endif /* regset.h */
   --- 51,62 ----
	function is COLLECT_REGSET.  If the regset has no collect function,
	pass NULL for COLLECT_REGSET.

   !    The object returned is allocated on ARCH's obstack.  */
   ! extern struct regset *(regset_obstack_alloc
   !                        (struct gdbarch *arch,
   !                         void *descr,
   !                         supply_regset_ftype *supply_regset,
   !                         collect_regset_ftype *collect_regset));



I'm not entirely happy with the name regset_obstack_alloc.  We have
frame_obstack_zalloc and gdbarch_obstack_zalloc, but that's because we
have a frame obstack and a gdbarch obstack.  There is no
regset_obstack, so I think regset_obstack_alloc is confusing.  How
about just regset_alloc?  The caller shouldn't really care about the
way the srorage is provided; the `struct gdbarch' argument already
ties it to a specific architecture.

Now that it gets tied to the architecture anyway, I wonder whether we
should get rid of the descr argument entirely.  I might want to tweak
that before you add register sets to any new targets...

Anyway, consider a patch with s/regset_obstack_alloc/regset_obstack/g
and gets rid of the funny parenthesis pre-approved.

Mark

P.S. Sorry to be so pedantic about function name, but I feel the lack
of consistency in naming function is causing too much confusion.