[PATCH 1/2] gdbserver/xtensa: drop xtensa_usrregs_info

[PATCH 1/2] gdbserver/xtensa: drop xtensa_usrregs_info

[Max Filippov]

xtensa_usrregs_info refers to undefined variables xtensa_num_regs and
xtensa_regmap. Drop xtensa_usrregs_info and replace pointer to usrregs
in regs_info with NULL since all registers are read/set through regsets.

2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
gdb/gdbserver/
	* linux-xtensa-low.c (xtensa_usrregs_info): remove.
	(regs_info): replace usrregs pointer with NULL.
---
 gdb/gdbserver/linux-xtensa-low.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/gdb/gdbserver/linux-xtensa-low.c b/gdb/gdbserver/linux-xtensa-low.c
index f7fafaf..e786da5 100644
--- a/gdb/gdbserver/linux-xtensa-low.c
+++ b/gdb/gdbserver/linux-xtensa-low.c
@@ -186,16 +186,10 @@ static struct regsets_info xtensa_regsets_info =
     NULL, /* disabled_regsets */
   };
 
-static struct usrregs_info xtensa_usrregs_info =
-  {
-    xtensa_num_regs,
-    xtensa_regmap,
-  };
-
 static struct regs_info regs_info =
   {
     NULL, /* regset_bitmap */
-    &xtensa_usrregs_info,
+    NULL, /* usrregs */
     &xtensa_regsets_info
   };
 
-- 
1.8.1.4

[PATCH 2/2] gdbserver/xtensa: fix typo in XCHAL_HAVE_LOOPS

[Max Filippov]

This fixes lbeg/lend/lcount registers handling through gdbserver.

2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
gdb/gdbserver/
	* linux-xtensa-low.c (xtensa_fill_gregset,
	xtensa_store_gregset): fix typo in XCHAL_HAVE_LOOPS
---
 gdb/gdbserver/linux-xtensa-low.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/gdb/gdbserver/linux-xtensa-low.c b/gdb/gdbserver/linux-xtensa-low.c
index e786da5..4daccee 100644
--- a/gdb/gdbserver/linux-xtensa-low.c
+++ b/gdb/gdbserver/linux-xtensa-low.c
@@ -59,7 +59,7 @@ xtensa_fill_gregset (struct regcache *regcache, void *buf)
 
   /* Loop registers, if hardware has it.  */
 
-#if XCHAL_HAVE_LOOP
+#if XCHAL_HAVE_LOOPS
   collect_register_by_name (regcache, "lbeg", (char*)&rset[R_LBEG]);
   collect_register_by_name (regcache, "lend", (char*)&rset[R_LEND]);
   collect_register_by_name (regcache, "lcount", (char*)&rset[R_LCOUNT]);
@@ -94,7 +94,7 @@ xtensa_store_gregset (struct regcache *regcache, const void *buf)
 
   /* Loop registers, if hardware has it.  */
 
-#if XCHAL_HAVE_LOOP
+#if XCHAL_HAVE_LOOPS
   supply_register_by_name (regcache, "lbeg", (char*)&rset[R_LBEG]);
   supply_register_by_name (regcache, "lend", (char*)&rset[R_LEND]);
   supply_register_by_name (regcache, "lcount", (char*)&rset[R_LCOUNT]);
-- 
1.8.1.4

[PATCH 0/2] Fixes for xtensa gdbserver

[Max Filippov]

Hi,

this series fixes gdbserver build for xtensa and zero overhead loop
registers handling.

Max Filippov (2):
  gdbserver/xtensa: drop xtensa_usrregs_info
  gdbserver/xtensa: fix typo in XCHAL_HAVE_LOOPS

 gdb/gdbserver/linux-xtensa-low.c | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

-- 
1.8.1.4

Re: [PATCH 2/2] gdbserver/xtensa: fix typo in XCHAL_HAVE_LOOPS

[Pedro Alves]

On 04/17/2015 02:14 AM, Max Filippov wrote:
> This fixes lbeg/lend/lcount registers handling through gdbserver.
> 
> 2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
> gdb/gdbserver/
> 	* linux-xtensa-low.c (xtensa_fill_gregset,
> 	xtensa_store_gregset): fix typo in XCHAL_HAVE_LOOPS

I was curious on whether it might have been that XCHAL_HAVE_LOOP
used to exist and was later renamed ("as in, how did this ever work
then?"), but googling around, it doesn't seem have ever existed
indeed.

Write instead:

2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
gdb/gdbserver/
	* linux-xtensa-low.c (xtensa_fill_gregset)
	(xtensa_store_gregset): Check XCHAL_HAVE_LOOPS instead
        of XCHAL_HAVE_LOOP.

OK with that change.

Thanks,
Pedro Alves

Re: [PATCH 1/2] gdbserver/xtensa: drop xtensa_usrregs_info

[Pedro Alves]

On 04/17/2015 02:14 AM, Max Filippov wrote:
> xtensa_usrregs_info refers to undefined variables xtensa_num_regs and
> xtensa_regmap. Drop xtensa_usrregs_info and replace pointer to usrregs
> in regs_info with NULL since all registers are read/set through regsets.

Eeek, that was added 2 years ago already...  :-P

> 
> 2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
> gdb/gdbserver/
> 	* linux-xtensa-low.c (xtensa_usrregs_info): remove.
> 	(regs_info): replace usrregs pointer with NULL.

Looks fine, thanks.

Thanks,
Pedro Alves

Re: [PATCH 2/2] gdbserver/xtensa: fix typo in XCHAL_HAVE_LOOPS

[Max Filippov]

On Fri, Apr 17, 2015 at 2:29 PM, Pedro Alves <palves@redhat.com> wrote:
> On 04/17/2015 02:14 AM, Max Filippov wrote:
>> This fixes lbeg/lend/lcount registers handling through gdbserver.
>>
>> 2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
>> gdb/gdbserver/
>>       * linux-xtensa-low.c (xtensa_fill_gregset,
>>       xtensa_store_gregset): fix typo in XCHAL_HAVE_LOOPS
>
> I was curious on whether it might have been that XCHAL_HAVE_LOOP
> used to exist and was later renamed ("as in, how did this ever work

It did work, only these three registers were inaccessible.

> then?"), but googling around, it doesn't seem have ever existed
> indeed.
>
> Write instead:
>
> 2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
> gdb/gdbserver/
>         * linux-xtensa-low.c (xtensa_fill_gregset)
>         (xtensa_store_gregset): Check XCHAL_HAVE_LOOPS instead
>         of XCHAL_HAVE_LOOP.
>
> OK with that change.

Ok. Can I check it in (I have write-after-approval access to binutils),
or do I need to request such access to gdb separately?

-- 
Thanks.
-- Max

Re: [PATCH 2/2] gdbserver/xtensa: fix typo in XCHAL_HAVE_LOOPS

[Pedro Alves]

On 04/17/2015 01:50 PM, Max Filippov wrote:
> On Fri, Apr 17, 2015 at 2:29 PM, Pedro Alves <palves@redhat.com> wrote:
>> On 04/17/2015 02:14 AM, Max Filippov wrote:
>>> This fixes lbeg/lend/lcount registers handling through gdbserver.
>>>
>>> 2015-04-17  Max Filippov  <jcmvbkbc@gmail.com>
>>> gdb/gdbserver/
>>>       * linux-xtensa-low.c (xtensa_fill_gregset,
>>>       xtensa_store_gregset): fix typo in XCHAL_HAVE_LOOPS
>>
>> I was curious on whether it might have been that XCHAL_HAVE_LOOP
>> used to exist and was later renamed ("as in, how did this ever work
> 
> It did work, only these three registers were inaccessible.

That's what I'd call "not work". :-)

(the point of that was that in case of a rename, it'd be
probably better to check both XCHAL_HAVE_LOOP/XCHAL_HAVE_LOOPS.)

>> OK with that change.
> 
> Ok. Can I check it in (I have write-after-approval access to binutils),
> or do I need to request such access to gdb separately?

Once the copyright issue is sorted out, please add yourself
to the gdb/MAINTAINERS file under Write After Approval.
(it's preapproved, just push it in and post it to the
gdb-patches@ list).

Since the fixes/patches are trivial, it's OK to push them
in before that is resolved.  Feel free to do that any time.

Thanks,
Pedro Alves