Re: [PATCH] gdb.trace: Fix string collection for 64-bit platforms.

["Marcin Kościelnicki" <koriakin@0x04.net>]

On 21/01/16 17:14, Pedro Alves wrote:
> On 01/21/2016 04:02 PM, Marcin Kościelnicki wrote:
>> String collection always used ref32 to fetch the string pointer.  Make it
>> use ref64 when pointeer size is 64-bit.
>
> "pointer".
>
>>
>> This appeared to work on x86_64 since it's a little-endian platform, and
>> malloc (used in gdb.trace/collection.exp) returns addresses in low 4GB.
>> Noticed and tested on s390x-ibm-linux-gnu, also tested on
>> i686-unknown-linux-gnu and x86_64-unknown-linux-gnu.
>>
>> gdb/ChangeLog:
>>
>> 	* ax-gdb.c (gen_traced_pop): Use aop_ref64 when appropriate for
>> 	string collection.
>> ---
>> OK to push?
>>
>>   gdb/ChangeLog | 5 +++++
>>   gdb/ax-gdb.c  | 5 ++++-
>>   2 files changed, 9 insertions(+), 1 deletion(-)
>>
>> diff --git a/gdb/ChangeLog b/gdb/ChangeLog
>> index 45d8ef9..3f4d152 100644
>> --- a/gdb/ChangeLog
>> +++ b/gdb/ChangeLog
>> @@ -1,3 +1,8 @@
>> +2016-01-21  Marcin Kościelnicki  <koriakin@0x04.net>
>> +
>> +	* ax-gdb.c (gen_traced_pop): Use aop_ref64 when appropriate for
>> +	string collection.
>> +
>>   2016-01-21  Andrew Burgess  <andrew.burgess@embecosm.com>
>>
>>   	* disasm.c (maybe_add_dis_line_entry): Rename to...
>> diff --git a/gdb/ax-gdb.c b/gdb/ax-gdb.c
>> index dd6eee6..d17cb04 100644
>> --- a/gdb/ax-gdb.c
>> +++ b/gdb/ax-gdb.c
>> @@ -410,7 +410,10 @@ gen_traced_pop (struct gdbarch *gdbarch,
>>
>>   	  if (string_trace)
>>   	    {
>> -	      ax_simple (ax, aop_ref32);
>> +	      if (TYPE_LENGTH(value->type) == 4)
>
> Missing space before parens.
>
>> +	        ax_simple (ax, aop_ref32);
>> +	      else
>> +	        ax_simple (ax, aop_ref64);
>
>> +	      if (TYPE_LENGTH(value->type) == 4)
>> +	        ax_simple (ax, aop_ref32);
>> +	      else
>> +	        ax_simple (ax, aop_ref64);
>
>
> Missing space before parens.  I was going to say to
> check for == 8 instead so that we don't pessimize the
> 8/16 bits cases, but sounds better even to factor out
> the switch in gen_fetch, so that the ref8/ref16 are covered
> as well:
>
>        switch (TYPE_LENGTH (type))
> 	{
> 	case 8 / TARGET_CHAR_BIT:
> 	  ax_simple (ax, aop_ref8);
> 	  break;
> 	case 16 / TARGET_CHAR_BIT:
> 	  ax_simple (ax, aop_ref16);
> 	  break;
> 	case 32 / TARGET_CHAR_BIT:
> 	  ax_simple (ax, aop_ref32);
> 	  break;
> 	case 64 / TARGET_CHAR_BIT:
> 	  ax_simple (ax, aop_ref64);
> 	  break;
>
> 	  /* Either our caller shouldn't have asked us to dereference
> 	     that pointer (other code's fault), or we're not
> 	     implementing something we should be (this code's fault).
> 	     In any case, it's a bug the user shouldn't see.  */
> 	default:
> 	  internal_error (__FILE__, __LINE__,
>
> WDYT?
>
> Thanks,
> Pedro Alves
>


Fair enough, ref16 may actually be useful on some tiny platforms, and it 
certainly won't hurt to throw in ref8 as well.  I'll do that.