Re: [PATCH 2/3] gdb: remove push_target free functions

[Simon Marchi <simon.marchi@polymtl.ca>]

On 2021-03-22 1:31 p.m., Simon Marchi via Gdb-patches wrote:
>>> index bcc54bd966fe..b737bcf6e0f3 100644
>>> --- a/gdb/exec.c
>>> +++ b/gdb/exec.c
>>> @@ -616,7 +616,7 @@ program_space::add_target_sections (void *owner,
>>>  	    continue;
>>>
>>>  	  switch_to_inferior_no_thread (inf);
>>> -	  push_target (&exec_ops);
>>> +	  inf->push_target (&exec_ops);
>>
>> Similar to the previous patch, it should be OK to also remove the 
>> switch_to_inferior_no_thread here.
> 
> Similarly, pushing a target can cause another (at the same stratum) to
> get unpushed, and some target_ops::close implementations refer to the
> current inferior.  So out of caution I would leave it.  I'll also add
> a note to the commit message here.

Here too, it should be safe to remove it, now that I think about it.
We check just before that the exec target isn't pushed on the inferior's
stack, and there isn't any other file_stratum that could be there.  So
I'll remove it.  Thanks for making me double check that.

>>> @@ -1183,7 +1183,7 @@ follow_exec (ptid_t ptid, const char *exec_file_target)
>>>
>>>        inferior *org_inferior = current_inferior ();
>>>        switch_to_inferior_no_thread (inf);
>>
>> And here.
> 
> Same.  Although here it's a brand new inferior, so we know there isn't
> an existing process target that is going to get unpushed.  But also
> here, is it maybe important for the current inferior to be correctly set
> for the remainder of the follow_exec function?

I'll leave this one, because I still think it's important for the
remainder of the follow_exec function.

Simon