Theodor Thornhill <theo@thornhill.no> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: Theodor Thornhill <theo@thornhill.no>
>>> Cc: 61558@debbugs.gnu.org
>>> Date: Fri, 17 Feb 2023 20:49:23 +0100
>>> 
>>> Can you test this patch?  It seems to work for me, but I'm no C expert.
>>
>> Thanks, this is much better.  However, the handling of #if (as opposed
>> to #ifdef) is still incorrect.  For example, go to the end of line
>> 3376 of dispnew.c and type RET: point goes to column zero, not column
>> 2.  If you convert "#if" to #ifdef", the behavior becomes correct.
>> Same problem exists with #elif.
>>
>> Doesn't tree-sitter grammar consider #if and #elif preprocessor nodes?
>
> It seems we are at the mercy of this[0] issue here, in all the cases
> you've described.
>
>>
>> Also, there's a problem with #define in the middle of code.  For
>> example, here:
>>
>>   /* Put a `display' property on the string for the captions to display,
>>      put a `menu_item' property on tab-bar items with a value that
>>      is the index of the item in F's tab-bar item vector.  */
>>   for (i = 0; i < f->n_tab_bar_items; ++i)
>>     {
>> #define PROP(IDX) \
>>   AREF (f->tab_bar_items, i * TAB_BAR_ITEM_NSLOTS + (IDX))
>>
>>       caption = Fcopy_sequence (PROP (TAB_BAR_ITEM_CAPTION));  <<<<<<<<<<
>>
>>       /* Put a `display' text property on the string for the caption to
>> 	 display.  Put a `menu-item' property on the string that gives
>> 	 the start of this item's properties in the tab-bar items
>> 	 vector.  */
>>       AUTO_LIST2 (props, Qmenu_item, make_fixnum (i * TAB_BAR_ITEM_NSLOTS));
>>
>>       Fadd_text_properties (make_fixnum (0), make_fixnum (SCHARS (caption)),
>> 			    props, caption);
>>
>>       f->desired_tab_bar_string =
>> 	concat2 (f->desired_tab_bar_string, caption);
>>
>> #undef PROP <<<<<<<<<<<<<<<<<<<<
>>     }
>>
>> Typing RET at the end of the two marked lines goes to column zero
>> instead of the expected non-zero column.  So it sounds like #define
>> and #undef are also not handled correctly yet.
>
> Yeah, luckily it indents correctly after you start to type.  I'll try to
> see if I can make some specific handling for this.
>
> Theo
>

Scratch that.  Can you test this for me?  I think I got it.

Theo