> Date: Sat, 10 Apr 2010 08:46:04 -0700 > From: "H.J. Lu" > > On Sat, Apr 10, 2010 at 7:49 AM, Mark Kettenis wrote: > >> Date: Fri, 9 Apr 2010 13:43:06 -0700 > >> From: "H.J. Lu" > >> > >> On Fri, Apr 9, 2010 at 1:08 PM, H.J. Lu wrote: > >> > On Fri, Apr 9, 2010 at 12:15 PM, Mark Kettenis wrote: > >> >>> Date: Fri, 9 Apr 2010 09:01:01 -0700 > >> >>> From: "H.J. Lu" > >> >>> > >> >>> On Fri, Apr 9, 2010 at 8:59 AM, Mark Kettenis wrote: > >> >>> >> Date: Fri, 9 Apr 2010 08:50:12 -0700 > >> >>> >> From: "H.J. Lu" > >> >>> >> > >> >>> >> Hi, > >> >>> >> > >> >>> >> This patch fixes gcore by not generating core regset sections which > >> >>> >> aren't supported by i386 without SSE.  OK to install? > >> >>> > > >> >>> > Sorry, I think this is the wrong way to fix this. > >> >>> > > >> >>> > >> >>> Any suggestions? > >> >> > >> >> Yup.  This makes sure we always install a list of register notes that > >> >> matches the target description. > >> >> > >> >> 2010-04-09  Mark Kettenis   > >> >> > >> >>        * i386-linux-tdep.c (i386_linux_regset_sections): Remove extended > >> >>        register note sections. > >> >>        (i386_linux_sse_regset_sections, i386_linux_avx_regset_sections): > >> >>        New variables. > >> >>        (i386_linux_init_abi): Install list of supported register note > >> >>        sections that matches the target description. > >> >> > >> > > >> > That is nice. We should do the same thing in amd64-linux-tdep.c. > >> > > >> > >> My second thought. No need to change amd64-linux-tdep.c. > >> We can just add i386_linux_mmx_regset_sections. OK to install? > > > > Why should we do it this way? > > > > It will be nice for "gcore" to generate all valid note sections > so that the older gdb can still exam the core generated by > the newer gdb even if the older gdb may not access the > new registers it doesn't support. Otherwise, gdb 7.0 > can't see any XMM registers in coredump generated > by gdb 7.2 on AVX machine even if the upper 128bits > of YMM registers are never used. Thought about this for a bit, but I think it is better not to include those registers at all than to generate core dumps where half the register appears to be "lost". So I committed my origional diff, and I'll change amd64-linux-tdep.c to match the i386 behaviour.