Printing to STDERR when %d is "Cannot access memory ..."

Printing to STDERR when %d is "Cannot access memory ..."

[Keith David Bershatsky]

I am working on crosshairs (17684) and multiple fake cursors (22873) and created a little problem for myself in that a character not recognized by the font (U0001D4D5) [aka a pretty letter f] on a line all by itself does not have a font->space_width and causes Emacs to crash when printing said value to STDERR.

GDB returns "Cannot access memory at address 0xa0" when printing the value of font->space_width.

This came about by using the following setting for a Form Feed character:

(setq buffer-display-table (make-display-table))
(aset buffer-display-table
      ?\014
      (vector (make-glyph-code ?\U0001D4D5 'font-lock-warning-face)))

I am by passing the problem by not using ?\U0001D4D5.  However, I would like to come up with programmatic solution to avoid crashing Emacs when printing values to STDERR.

I tried:

fprintf (stderr, "(%d)", (font->space_width == NULL
                          ? 0
                          : font->space_width);

However, that didn't fix the problem.

Thanks,

Keith

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Davis Herring]

> fprintf (stderr, "(%d)", (font->space_width == NULL
>                            ? 0
>                            : font->space_width);
> 
> However, that didn't fix the problem.

You meant

   fprintf (stderr, "(%d)", font == NULL ? 0 :
                            font->space_width);

It's font that could be null, not its space_width.

Davis

PS - This says nothing about why font is null -- it might be a bug, or 
it might just be a bug that this code uses it without checking.

-- 
This product is sold by volume, not by mass.  If it appears too dense or 
too sparse, it is because mass-energy conversion has occurred during 
shipping.

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Eli Zaretskii]

> Date: Wed, 13 Jun 2018 11:37:09 -0700
> From: Keith David Bershatsky <esq@lawlist.com>
> 
> I am working on crosshairs (17684) and multiple fake cursors (22873) and created a little problem for myself in that a character not recognized by the font (U0001D4D5) [aka a pretty letter f] on a line all by itself does not have a font->space_width and causes Emacs to crash when printing said value to STDERR.
> 
> GDB returns "Cannot access memory at address 0xa0" when printing the value of font->space_width.

This message means that 'font' is a NULL pointer, and that is your
basic problem.  You need to find out why 'font' is NULL.

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Keith David Bershatsky]

Thank you, Davis, for teaching me how to deal with a NULL pointer when printing to STDERR.  Your code works well.  Greatly appreciated!

Keith

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

DATE:  [06-13-2018 11:50:13] <13 Jun 2018 12:50:13 -0600>
FROM:  Davis Herring <herring@lanl.gov>
> 
> > fprintf (stderr, "(%d)", (font->space_width == NULL
> >                            ? 0
> >                            : font->space_width);
> >
> > However, that didn't fix the problem.
> 
> You meant
> 
>    fprintf (stderr, "(%d)", font == NULL ? 0 :
>                             font->space_width);
> 
> It's font that could be null, not its space_width.
> 
> Davis
> 
> PS - This says nothing about why font is null -- it might be a bug, or
> it might just be a bug that this code uses it without checking.
> 
> --
> This product is sold by volume, not by mass.  If it appears too dense or
> too sparse, it is because mass-energy conversion has occurred during
> shipping.

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Keith David Bershatsky]

Thank you, Eli, for taking a look at this thread.  The suggestion by Davis to deal with the STDERR message avoids the crash.

My STDERR printout (with the NULL font now being assigned a 0 value) is as follows:

    it->c (120021)
    char (ð)
    w->hscroll (0)
    hscl (false)
    it->hpos (5)
    it->vpos (0)
    it->current_x (55)
    new_x (55)
    font->space_width (0)
    it->lnum_pixel_width (55)
    it->lnum_width (3)
    it->first_visible_x (0)
    relative_x (55)
    it->face_id (38)
    foreground_face (magenta)
    background_face (black)
    rc (MOVE_X_REACHED)

I create the font pointer like this ....  It has worked for everything so far except this special character (U0001D4D5) on Emacs built --with-x.  I believe the face_id is correct in this example because the foreground (magenta) and background (black) are correctly returned.

  struct face *face = mc_lookup_face (it);
  struct font *font = face->font;
  Lisp_Object foreground_face = face->lface[LFACE_FOREGROUND_INDEX];
  Lisp_Object background_face = face->lface[LFACE_BACKGROUND_INDEX];

... using the following function:

struct face *
mc_lookup_face (struct it *it)
{
  if (!NILP (Vface_remapping_alist)
      && (it->face_id == DEFAULT_FACE_ID /* Qdefault */
          || it->face_id == MODE_LINE_FACE_ID /* Qmode_line */
          || it->face_id == MODE_LINE_INACTIVE_FACE_ID /* Qmode_line_inactive */
          || it->face_id == HEADER_LINE_FACE_ID /* Qheader_line */
          || it->face_id == TOOL_BAR_FACE_ID /* Qtool_bar */
          || it->face_id == FRINGE_FACE_ID /* Qfringe */
          || it->face_id == SCROLL_BAR_FACE_ID /* Qscroll_bar */
          || it->face_id == BORDER_FACE_ID /* Qborder */
          || it->face_id == CURSOR_FACE_ID /* Qcursor */
          || it->face_id == MOUSE_FACE_ID /* Qmouse */
          || it->face_id == MENU_FACE_ID /* Qmenu */
          || it->face_id == WINDOW_DIVIDER_FACE_ID /* Qwindow_divider */
          || it->face_id == WINDOW_DIVIDER_FIRST_PIXEL_FACE_ID /* Qwindow_divider_first_pixel */
          || it->face_id == WINDOW_DIVIDER_LAST_PIXEL_FACE_ID /* Qwindow_divider_last_pixel */
          || it->face_id == VERTICAL_BORDER_FACE_ID /* Qvertical_border */
          || it->face_id == INTERNAL_BORDER_FACE_ID)) /* Qinternal_border */
    return FACE_FROM_ID (it->f, lookup_basic_face (it->f, it->face_id));
      else
        return FACE_FROM_ID (it->f, it->face_id);
}

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

DATE:  [06-13-2018 12:11:33] <13 Jun 2018 22:11:33 +0300>
FROM:  Eli Zaretskii <eliz@gnu.org>
> 
> > Date: Wed, 13 Jun 2018 11:37:09 -0700
> > From: Keith David Bershatsky <esq@lawlist.com>
> >
> > I am working on crosshairs (17684) and multiple fake cursors (22873) and created a little problem for myself in that a character not recognized by the font (U0001D4D5) [aka a pretty letter f] on a line all by itself does not have a font->space_width and causes Emacs to crash when printing said value to STDERR.
> >
> > GDB returns "Cannot access memory at address 0xa0" when printing the value of font->space_width.
> 
> This message means that 'font' is a NULL pointer, and that is your
> basic problem.  You need to find out why 'font' is NULL.

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Eli Zaretskii]

> Date:  Wed, 13 Jun 2018 12:34:14 -0700
> From:  Keith David Bershatsky <esq@lawlist.com>
> Cc:  emacs-devel@gnu.org,Davis Herring <herring@lanl.gov>
> 
> My STDERR printout (with the NULL font now being assigned a 0 value) is as follows:
> 
>     it->c (120021)
>     char (ð)
>     w->hscroll (0)
>     hscl (false)
>     it->hpos (5)
>     it->vpos (0)
>     it->current_x (55)
>     new_x (55)
>     font->space_width (0)
>     it->lnum_pixel_width (55)
>     it->lnum_width (3)
>     it->first_visible_x (0)
>     relative_x (55)
>     it->face_id (38)
>     foreground_face (magenta)
>     background_face (black)
>     rc (MOVE_X_REACHED)
> 
> I create the font pointer like this ....  It has worked for everything so far except this special character (U0001D4D5) on Emacs built --with-x.  I believe the face_id is correct in this example because the foreground (magenta) and background (black) are correctly returned.
> 
>   struct face *face = mc_lookup_face (it);
>   struct font *font = face->font;
>   Lisp_Object foreground_face = face->lface[LFACE_FOREGROUND_INDEX];
>   Lisp_Object background_face = face->lface[LFACE_BACKGROUND_INDEX];
> 
> ... using the following function:
> 
> struct face *
> mc_lookup_face (struct it *it)
> {
>   if (!NILP (Vface_remapping_alist)
>       && (it->face_id == DEFAULT_FACE_ID /* Qdefault */
>           || it->face_id == MODE_LINE_FACE_ID /* Qmode_line */
>           || it->face_id == MODE_LINE_INACTIVE_FACE_ID /* Qmode_line_inactive */
>           || it->face_id == HEADER_LINE_FACE_ID /* Qheader_line */
>           || it->face_id == TOOL_BAR_FACE_ID /* Qtool_bar */
>           || it->face_id == FRINGE_FACE_ID /* Qfringe */
>           || it->face_id == SCROLL_BAR_FACE_ID /* Qscroll_bar */
>           || it->face_id == BORDER_FACE_ID /* Qborder */
>           || it->face_id == CURSOR_FACE_ID /* Qcursor */
>           || it->face_id == MOUSE_FACE_ID /* Qmouse */
>           || it->face_id == MENU_FACE_ID /* Qmenu */
>           || it->face_id == WINDOW_DIVIDER_FACE_ID /* Qwindow_divider */
>           || it->face_id == WINDOW_DIVIDER_FIRST_PIXEL_FACE_ID /* Qwindow_divider_first_pixel */
>           || it->face_id == WINDOW_DIVIDER_LAST_PIXEL_FACE_ID /* Qwindow_divider_last_pixel */
>           || it->face_id == VERTICAL_BORDER_FACE_ID /* Qvertical_border */
>           || it->face_id == INTERNAL_BORDER_FACE_ID)) /* Qinternal_border */
>     return FACE_FROM_ID (it->f, lookup_basic_face (it->f, it->face_id));
>       else
>         return FACE_FROM_ID (it->f, it->face_id);
> }

I don't think face->font is guaranteed to be non-NULL.  You need to
check that explicitly, and you need some fallback for when it is NULL,
to serve as a substitute for font->space_width.

Re: Printing to STDERR when %d is "Cannot access memory ..."

[Keith David Bershatsky]

Thank you, Eli.  It is a good thing that I asked about this issue because a few of my calculations require a non-NULL value.  I will be using FRAME_COLUMN_WIDTH (f) as the fallback method when font is NULL, which will probably suffice for this particular project.

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;

DATE:  [06-13-2018 12:46:54] <13 Jun 2018 22:46:54 +0300>
FROM:  Eli Zaretskii <eliz@gnu.org>
> 
> * * *
> 
> I don't think face->font is guaranteed to be non-NULL.  You need to
> check that explicitly, and you need some fallback for when it is NULL,
> to serve as a substitute for font->space_width.