bug#36019: 27.0.50; Segfault in color_distance in emacs --batch

[Alex Gramiak <agrambot@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1036 bytes --]

Vitalie Spinu <spinuvit@gmail.com> writes:

> I have bisected it to:
>
> yes: 5d8b0fadee * | Add terminal hook query_frame_background_color
> err: 41e20ee4bc * | Add terminal hook defined_color_hook
> err: a411517faf * | Rename generic x_* identifiers
> err: da9541dd10 * | Add prefixes to some window system-dependent procedures
>  no: ff4e31fa32 * |  Rename generic x_* procedures in xdisp.c
>
> The commits marked with err: don't build. So should be one of the top 4 commits.
>
>
>   Vitalie
>
>>> On Fri, May 31 2019 10:30, Vitalie Spinu wrote:
>
>> Hi,
>
>> One of my packages started segfaulting in tests after emacs 27 update. I cannot
>> reproduce it in an X session, nor make a simple reproducible example. Hopefully
>> the backtraces will be enough.
>
>> The issue appeared in the last 4-5 months. I can bisect if needed.

I can't reproduce this, but looking at the likely culprit (41e20ee4bc)
it's likely due to the frame being the initial frame. Just to make sure,
could you test with this simple primitive check for this?


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: color_distance.diff --]
[-- Type: text/x-patch, Size: 568 bytes --]

diff --git a/src/xfaces.c b/src/xfaces.c
index d211ec8c46..10cfa50526 100644
--- a/src/xfaces.c
+++ b/src/xfaces.c
@@ -4207,6 +4207,11 @@ two lists of the form (RED GREEN BLUE) aforementioned. */)
   struct frame *f = decode_live_frame (frame);
   Emacs_Color cdef1, cdef2;
 
+  fprintf (stderr,
+           "Initial: %d, hook: %p",
+           FRAME_INITIAL_P (f),
+           FRAME_TERMINAL (f)->defined_color_hook);
+
   if (!(CONSP (color1) && parse_rgb_list (color1, &cdef1))
       && !(STRINGP (color1)
            && FRAME_TERMINAL (f)->defined_color_hook (f,

[-- Attachment #3: Type: text/plain, Size: 302 bytes --]


I expect that it will print "Initial: 1, hook: 0x0" before segfaulting.

If this is the case, then perhaps the initial frame should also have
this hook defined, though this would be the first such hook defined for
the initial frame type.

If this is correct, then the following diff should fix this:


[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #4: tty_defined_color --]
[-- Type: text/x-patch, Size: 510 bytes --]

diff --git a/src/terminal.c b/src/terminal.c
index 0ee0121e35..ed2cfead7a 100644
--- a/src/terminal.c
+++ b/src/terminal.c
@@ -624,6 +624,7 @@ init_initial_terminal (void)
   initial_terminal->kboard = initial_kboard;
   initial_terminal->delete_terminal_hook = &delete_initial_terminal;
   initial_terminal->delete_frame_hook = &initial_free_frame_resources;
+  initial_terminal->defined_color_hook = &tty_defined_color; /* xfaces.c */
   /* Other hooks are NULL by default.  */
 
   return initial_terminal;