[ was: Re: [PATCH][gdb] Make INTERP_TUI's default ui_out the
INTERP_CONSOLE ui_out ]

On 01-10-19 14:59, Tom de Vries wrote:
> On 25-09-19 10:36, Andrew Burgess wrote:
>> * Tom de Vries <tdevries@suse.de> [2019-09-25 01:20:22 +0200]:
>>
>>> On 24-09-19 19:31, Andrew Burgess wrote:
>>>> With this patch applied I see the fixed behaviour you describe at the
>>>> CLI, however, if I do:
>>>>
>>>>    (gdb) define mybt
>>>>    bt
>>>>    end
>>>>    (gdb) start
>>>>    (gdb) set logging redirect on
>>>>    (gdb) tui enable
>>>>    (gdb) set logging on
>>>>    (gdb) mybt
>>>>
>>>> The I see output appear on both the console and in the log file.
>>>
>>> I can't reproduce this. Are you sure you removed the gdb.txt file before
>>> starting gdb?
>>
>> You're absolutely right, nothing goes to the log for `mybt`, this was
>> a silly mistake on my side.  Apologies.  However...
>>
>> With tui enabled and logging on and redirect on, the 'bt' __does__
>> write to the log, while 'mybt' writes to the screen.  This is the same
>> bug you're fixing for non-tui mode, correct?
> 
> From user perspective, yes.
> 
>> Any solution should
>> ideally address both cases, or at least be accompanied with an
>> explanation for why these problems are distinct and should be solved
>> separately.
>>
> 
> Ideally yes, agreed.
> 
> This tentative patch fixes both cases, by limiting the fix of commit
> 3a87ae656c28 "Use console uiout when executing breakpoint commands" to
> the case for which the corresponding problem was reported: INTERP_MI. [
> Adding CC simark ]
> 
> At this point I don't understand yet why special-casing INTERP_MI would
> be the correct thing to do here.

I've investigated a bit further the problem that commit 3a87ae656c28
fixes, and constructed a hopefully convincing rationale.

OK for trunk?

Thanks,
- Tom