Re: [patch v4 02/13] thread, btrace: add generic branch trace support

[Pedro Alves <palves@redhat.com>]

[Sorry, I had to rush out in a hurry before, and didn't trim the previous
reply.  I  was just having trouble myself finding the real reply among all
the quoting, so here it is again:]

>>
>>       /* The first block ends at the current pc.  */
>>       if (!VEC_empty (btrace_block_s, btp->btrace))
>> 	{
>> 	  struct frame_info *frame = get_current_frame ();
>
> This get_current_frame call here looks fishy.  This function takes a
> btrace_thread_info, and its callers work with a thread_info directly,
> which indicates that they may work with some current thread other than
> the thread passed in as argument.
>
>>
>> 	  if (frame)
>> 	    {
>
> What's this check supposed to mean?  get_current_frame never
> returns NULL.
>
>> 	      struct btrace_block *head =
>> 		VEC_index (btrace_block_s, btp->btrace, 0);
>
>>
>> 	      if (head && !head->end)
>> 		head->end = get_frame_pc (frame);

I meant to suggest bypassing frames, and using regcache_read_pc instead.

-- 
Pedro Alves