Re: [PATCH 3/3] GDB: New target s12z

[Simon Marchi <simark at simark dot ca>]

On 2018-08-26 2:15 p.m., Simon Marchi wrote:
> On 2018-08-26 1:41 p.m., John Darrington wrote:
>> Hi Simon,
>>
>> Thanks for the review.
>>
>> On Sun, Aug 26, 2018 at 01:19:29PM -0400, Simon Marchi wrote:
>>      Hi John,
>>      
>>      I did a first pass review and noted a few minor nits.  I didn't get too deep in the
>>      frame_id/unwind code because I'm not too familiar with that.
>>
>> Pity.  I was hoping someone could help me there :{P
> 
> I just haven't written enough of those to be able to spot bugs off-hand in them.
> 
> Maybe just one insight, it seems (I'm not sure) that s12z_frame_cache uses the SP value of the
> current frame (the one for which we compute the id) in the frame id.
> 
> The usual thing to do (I looked at a few other arches) is to use the same value as the
> canonical frame address as defined by DWARF (Section 6.4 in DWARF5.pdf), which is the
> value of the stack pointer just before that frame was created.
> 
> This is of course not mandatory, but I suppose that adhering to this de facto rule could
> make things clearer.
> 
>> Many of your comments and questions stem from the code which I used as a
>> pattern, so I'm unsure of the answers.  But I will find out as best I
>> can and prepare a new patch within the next few days.
>>
>> J'
> 
> Thanks,
> 
> Simon

I also noticed that the gdb.base/all-architectures test fails when testing s12z.  Try to run

  make check TESTS="gdb.base/all-architectures-*.exp"

in the gdb build directory, see if you get a bunch of:

  FAIL: gdb.base/all-architectures-6.exp: tests: osabi=AIX: arch=s12z: endian=auto: disassemble 0x0,+4 (GDB internal error)

Simon