bug#44047: current HEAD has been crashing on startup

[Eli Zaretskii]

Ping!  Any news on this?

> Date: Wed, 21 Oct 2020 19:03:20 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 44047@debbugs.gnu.org
> 
> > Date: Tue, 20 Oct 2020 16:19:26 -0600 (MDT)
> > From: Liam Stitt <stittl@cuug.ab.ca>
> > cc: 44047@debbugs.gnu.org
> > 
> > (gdb) frame 1
> > #1  0x0000555555597de8 in build_frame_matrix_from_leaf_window (
> >      frame_matrix=0x555555d4a2b0, w=0x555555d42420) at dispnew.c:2531
> > 2531          fill_up_frame_row_with_spaces (frame_row, 
> > window_matrix->matrix_x);
> > (gdb) print frame_matrix->rows
> > $1 = (struct glyph_row *) 0x0
> > (gdb) print frame_matrix->nrows
> > $2 = 0
> 
> OK, so this is the problem: the frame glyph matrix has no glyph rows
> allocated to it.  The question is how did that happen?
> 
> To find out, please run Emacs under GDB, after setting a breakpoint in
> adjust_frame_glyphs_for_frame_redisplay:
> 
>   (gdb) break adjust_frame_glyphs_for_frame_redisplay
>   (gdb) run -Q -nw
> 
> When the breakpoint breaks, step through the code with the "next"
> command.  The first couple of times when the function is called, it
> should return immediately, here:
> 
>   if (!FRAME_LIVE_P (f))
>     return;
> 
> In the call after those 2, it should allocate the glyph pools and the
> 2 glyph matrices, then call allocate_matrices_for_frame_redisplay, and
> return here:
> 
>         if (!FRAME_WINDOW_P (f) && pool_changed_p)
>           SET_FRAME_GARBAGED (f);
>         return;
> 
> And in the next call, it should in addition allocate the glyph rows,
> here:
> 
>       else
>       {
>         adjust_glyph_matrix (NULL, f->desired_matrix, 0, 0, matrix_dim);
>         adjust_glyph_matrix (NULL, f->current_matrix, 0, 0, matrix_dim);
>         SET_FRAME_GARBAGED (f);
>       }
> 
> This last call's backtrace should look like this:
> 
>   (gdb) bt
>   #0  adjust_frame_glyphs_for_frame_redisplay (f=0xf46948) at dispnew.c:2008
>   #1  0x000000000041c905 in adjust_frame_glyphs (f=0xf46948) at dispnew.c:1828
>   #2  0x000000000042cdfd in adjust_frame_size (f=0xf46948, new_width=80,
>       new_height=37, inhibit=5, pretend=false, parameter=XIL(0x3690))
>       at frame.c:819
>   #3  0x0000000000427657 in change_frame_size_1 (f=0xf46948, new_width=80,
>       new_height=37, pretend=false, delay=false, safe=true, pixelwise=false)
>       at dispnew.c:5798
>   #4  0x00000000004276aa in change_frame_size (f=0xf46948, new_width=80,
>       new_height=37, pretend=false, delay=false, safe=true, pixelwise=false)
>       at dispnew.c:5830
>   #5  0x00000000004290e2 in init_display_interactive () at dispnew.c:6397
>   #6  0x0000000000429561 in init_display () at dispnew.c:6449
>   #7  0x00000000005a892b in main (argc=3, argv=0x7fffffffe538) at emacs.c:1995
> 
> Does this happen to you as I described, or does something different
> happen?
> 
> Also, each time allocate_matrices_for_frame_redisplay is called,
> please show the value of matrix_dim after the call:
> 
>   (gdb) p matrix_dim
> 
> and also show the value of f->total_lines each time
> adjust_frame_glyphs_for_frame_redisplay is called:
> 
>   (gdb) p f->total_lines.
> 
> Finally, the last time adjust_frame_glyphs_for_frame_redisplay is
> called, and goes on through all of its code, you should see non-NULL
> value of the frame's glyph_matrix's glyph_rows before the function
> returns.  Here's what I see:
> 
>   (gdb) p f->desired_matrix->rows
>   $11 = (struct glyph_row *) 0xf50140
>   (gdb) p f->desired_matrix->nrows
>   $12 = 38
> 
> (In my case, the TTY window has height of 38 lines.)
> 
> If what happens on your system matches the above description, then
> somehow the value of f->desired_matrix->rows is NULLified after this
> initialization.  But if you get NULL as f->desired_matrix->rows in the
> last call to adjust_frame_glyphs_for_frame_redisplay, or if Emacs
> segfaults before it reaches that point, then something abnormal
> happens during initialization, and we will next look at that.
> 
> Thanks.
> 
> 
> 
>