Re: [PATCH] RFC: gdb: redo whitespace-stripping from commands

[Guinevere Larsen <guinevere@redhat.com>]

On 5/29/25 10:36 AM, Tom Tromey wrote:
>>>>>> "Kartik" == Kartik K Agaram <ak@akkartik.com> writes:
> Thanks for the patch.
>
> Kartik> Before this patch, trailing whitespace was not stripped:
> Kartik>   - from 'set'/'show' commands
> Kartik>   - from 'complete' commands
>
> Kartik> Now I've added the 'with' commands to that list because it contains a
> Kartik> 'complete' subcommand. To accomplish this, I'm trying to fix a TODO to
> Kartik> provide a per-command flag controlling whitespace-stripping.
>
> Kartik> I've also cleaned up some seemingly TODOs that are either fixed by this
> Kartik> patch or obsolete.
>
> Kartik> Open question: This patch stops stripping trailing whitespace from all
> Kartik> 'with' commands. Might that create issues with other 'with' commands
> Kartik> besides 'with ... -- complete'?
>
> Maybe it would cause problems with trailing whitespace showing up in
> filenames?  Or with commands that don't expect an argument.  For those
> commands, completion might add a trailing space, causing the function to
> be called with " " rather than nullptr.
>
> I wonder if the 'with' command itself -- and any other commands that
> invoke commands, like 'eval' -- should strip whitespace as appropriate
> before invocation.  That is, move the whitespace stripping code
> somewhere else.  (Assuming it's in the wrong spot, I don't recall off
> the top of my head where this stuff happens.)

Removing trailing whitespace is done in the execute_command function in 
top.c. The command "with" will already call execute_command itself to 
invoke the correct command, so I think the correct way to go about it is 
for those commands to never strip whitespace, and then always call 
execute_command, which will do the right thing based on the inner command.

>
> Kartik> +  complete_cmd->strip_trailing_white_space_p = 0;
>   
> Kartik> +  c->strip_trailing_white_space_p = 0;
>   
> Kartik> +      strip_trailing_white_space_p (1),
>
> Kartik> +  unsigned int strip_trailing_white_space_p : 1;
>
> This should be bool and the constants above should be true/false.
Should the name end in _p? I would think _p is for pointers, and this is 
just a bool, but maybe I'm mistaken.

-- 
Cheers,
Guinevere Larsen
She/Her/Hers

>
> Tom
>