bug#35814: 27.0.50; Broken font display

bug#35814: 27.0.50; Broken font display

[Juri Linkov]

[-- Attachment #1: Type: text/plain, Size: 336 bytes --]

Some yesterday's commit in master broke the display of the text that uses
the font "-misc-fixed-medium-r-normal--10-*-*-*-c-60-iso10646-*"
in GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.22.30)
Windowing system distributor 'The X.Org Foundation', version 11.0.11906000
System Description: Linux Mint 19

Screenshot:


[-- Attachment #2: 20190520.png --]
[-- Type: image/png, Size: 3484 bytes --]

bug#35814: 27.0.50; Broken font display

[Alex Gramiak]

Juri Linkov <juri@linkov.net> writes:

> Some yesterday's commit in master broke the display of the text that uses
> the font "-misc-fixed-medium-r-normal--10-*-*-*-c-60-iso10646-*"
> in GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.22.30)
> Windowing system distributor 'The X.Org Foundation', version 11.0.11906000
> System Description: Linux Mint 19
>
> Screenshot:

It's probably due to my a4fe9c70af7 commit, sorry.

I tested with all X font backends -- what's the value of
(frame-parameter nil 'font-backend) normally?

And is it only that font, or other fonts as well?

bug#35814: 27.0.50; Broken font display

[Katsumi Yamaoka]

On Tue, 21 May 2019 00:01:39 -0600, Alex Gramiak wrote:
> It's probably due to my a4fe9c70af7 commit, sorry.

> I tested with all X font backends -- what's the value of
> (frame-parameter nil 'font-backend) normally?

> And is it only that font, or other fonts as well?

It happens to me too with the japanese-jisx0208 font:

"-misc-fixed-medium-r-normal-*-16-*-*-*-*-*-jisx0208.1983-0"

After starting emacs -Q the `C-h h' display is all ok, however
setting that font breaks the display for the cjk text.

(frame-parameter nil 'font-backend)
 => (xft x)

In GNU Emacs 27.0.50 (build 1, x86_64-pc-cygwin, GTK+ Version 3.22.28)
 of 2019-05-21 built on localhost

Regards,

bug#35814: 27.0.50; Broken font display

[Andy Moreton]

On Tue 21 May 2019, Alex Gramiak wrote:

> Juri Linkov <juri@linkov.net> writes:
>
>> Some yesterday's commit in master broke the display of the text that uses
>> the font "-misc-fixed-medium-r-normal--10-*-*-*-c-60-iso10646-*"
>> in GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.22.30)
>> Windowing system distributor 'The X.Org Foundation', version 11.0.11906000
>> System Description: Linux Mint 19
>>
>> Screenshot:
>
> It's probably due to my a4fe9c70af7 commit, sorry.
>
> I tested with all X font backends -- what's the value of
> (frame-parameter nil 'font-backend) normally?
>
> And is it only that font, or other fonts as well?

Alex, at the end of xfont_draw() in xfont.c, we now have:

      else
        {
          const unsigned code = s->char2b[from];
          const XChar2b char2b = { .byte1 = code >> 8,
                                   .byte2 = code & 0xFF };
          XDrawImageString16 (display, FRAME_X_DRAWABLE (s->f),
                              gc, x, y, &char2b, len);
        }

and:
      else
        {
          const unsigned code = s->char2b[from];
          const XChar2b char2b = { .byte1 = code >> 8,
                                   .byte2 = code & 0xFF };
          XDrawString16 (display, FRAME_X_DRAWABLE (s->f),
                         gc, x, y, &char2b, len);
        }

Both of these pass "len", but use a single XChar2b on the stack, so
this causes an overrun (and draws random stack contents).

    AndyM

bug#35814: 27.0.50; Broken font display

[Juri Linkov]

>> Some yesterday's commit in master broke the display of the text that uses
>> the font "-misc-fixed-medium-r-normal--10-*-*-*-c-60-iso10646-*"
>> in GNU Emacs 27.0.50 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.22.30)
>> Windowing system distributor 'The X.Org Foundation', version 11.0.11906000
>> System Description: Linux Mint 19
>>
>> Screenshot:
>
> It's probably due to my a4fe9c70af7 commit, sorry.

Yes, after reverting a4fe9c70af7 locally,
everything is fine as before.

> I tested with all X font backends -- what's the value of
> (frame-parameter nil 'font-backend) normally?

(frame-parameter nil 'font-backend)
=> (xft x)

> And is it only that font, or other fonts as well?

I don't know about other fonts.  At least, I see that
the default font was fine.  And as you can see on the
previous screenshot of the etc/TODO file, parts of the
buffer that display text with the italic face, are
not affected by this problem.

bug#35814: 27.0.50; Broken font display

[Alex Gramiak]

Andy Moreton <andrewjmoreton@gmail.com> writes:

> Alex, at the end of xfont_draw() in xfont.c, we now have:
>
>       else
>         {
>           const unsigned code = s->char2b[from];
>           const XChar2b char2b = { .byte1 = code >> 8,
>                                    .byte2 = code & 0xFF };
>           XDrawImageString16 (display, FRAME_X_DRAWABLE (s->f),
>                               gc, x, y, &char2b, len);
>         }
>
> and:
>       else
>         {
>           const unsigned code = s->char2b[from];
>           const XChar2b char2b = { .byte1 = code >> 8,
>                                    .byte2 = code & 0xFF };
>           XDrawString16 (display, FRAME_X_DRAWABLE (s->f),
>                          gc, x, y, &char2b, len);
>         }
>
> Both of these pass "len", but use a single XChar2b on the stack, so
> this causes an overrun (and draws random stack contents).

Wow, sorry, that was a bad mistake. I've applied commit 5a024b72c to
master; can those who experience the bug affirm that it's been fixed?

It appears that I only tested with fonts that make s->padding_p true,
and I can't seem to find a font on my system that doesn't.

bug#35814: 27.0.50; Broken font display

[Juri Linkov]

> Wow, sorry, that was a bad mistake. I've applied commit 5a024b72c to
> master; can those who experience the bug affirm that it's been fixed?
>
> It appears that I only tested with fonts that make s->padding_p true,
> and I can't seem to find a font on my system that doesn't.

Thanks!  Now everything is back to normal.  Closing.