bug#61558: 29.0.60; Indentation with c-ts-mode doesn't work in code guarded by #ifdef..#endif

[Theodor Thornhill via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

Eli Zaretskii <eliz@gnu.org> writes:

> To reproduce:
>
>   emacs -Q
>   C-x C-f src/dispnew.c RET
>   C-u 265 M-g g
>   C-e
>   RET
>
> The code around there looks like this:
>
> static struct glyph_matrix *
> new_glyph_matrix (struct glyph_pool *pool)
> {
>   struct glyph_matrix *result = xzalloc (sizeof *result);
>
> #if defined GLYPH_DEBUG && defined ENABLE_CHECKING
>   /* Increment number of allocated matrices.  This count is used
>      to detect memory leaks.  */
>   ++glyph_matrix_count;  <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
> #endif
>
> Line 265 is the one indicated with "<<<<<<".  Pressing RET goes to
> column 0, not column 2 as expected.  It looks like indentation doesn't
> work in code fragments that are guarded by #ifdef..#endif preprocessor
> conditions.  I tried several such places, for example, lines 295, 796,
> 1338, 1360, 1745, 2401, 2615, and many more.
>

Yep. We need rules for these in particular.  They aren't really
straightforward because the expected indent style varies, afaict.  For
example:

#ifdef WINDOWSNT
#include "w32.h"
#endif

compared to

#if defined GLYPH_DEBUG && defined ENABLE_CHECKING
  /* Increment number of allocated matrices.  This count is used
     to detect memory leaks.  */
  ++glyph_matrix_count;
#endif


Is it a correct assuption to think that whatever is inside one of these
if-blocks should indent according to their grand-parents rule?

In this case:


static struct glyph_matrix *
new_glyph_matrix (struct glyph_pool *pool)
{
  struct glyph_matrix *result = xzalloc (sizeof *result);

#if defined GLYPH_DEBUG && defined ENABLE_CHECKING
  /* Increment number of allocated matrices.  This count is used
     to detect memory leaks.  */
  ++glyph_matrix_count;
#endif

  /* Set pool and return.  */
  result->pool = pool;
  return result;
}

  ++glyph_matrix_count;

is indented one step from the compound_statement node, right?  So we
need a way to "ignore" the parents indentation.


> Strangely, in other places indentation does work: lines 1069, 3119.
>

Yeah, in these cases we have something other than the preproc directive
itself to indent from.

Theo