mintty overstrokes some fonts unexpectedly

mintty overstrokes some fonts unexpectedly

[Lemures Lemniscati]

Hi!

mintty overstrokes some fonts unexpectedly.
https://gitlab.com/test.cases/mintty-test/-/tree/54ae800e695ecd1741851cab57320a9d0e95a6fd


How to reproduce

Setting mintty to use msgothic.ttc in 20pt, run a
following command.

====
perl -e '
binmode STDOUT, ":utf8";
print <<END;
01234567
\x{3042}\x{3044}\x{3046}\x{3048}
AaBbCcDe
\x{25cb}\x{25cb}\x{25cb}\x{25cb}
\x{3007}\x{3007}\x{3007}\x{3007}
\x{300c}\x{300e}\x{3010}\x{3012}
\x{300d}\x{300f}\x{3011}\x{3013}
\x{ff62}\x{ff63}\x{ff62}\x{ff63}\x{ff62}\x{ff63}\x{ff62}\x{ff63}
\x{ff71}\x{ff72}\x{ff73}\x{ff74}\x{ff75}\x{ff76}\x{ff77}\x{ff78}
END
'
====


I got a result mintty-sample-msgothic.png.
https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/mintty-sample-msgothic.png

In the 4th line of the output, fonts (of u+25cb) were overstruck
unexpectedly.
And there are other characters also, which are similarly overstruck.


On the other hand, copying the result to a window of notepad.exe
displaying with msgothic.ttc in 20pt.

I got a result notepad-sample-msgothic.png, in which fonts were rendered
expectedly.
https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/notepad-sample-msgothic.png


My cygwin environment is:

====
$ uname -srvmo; cygcheck -c mintty;
CYGWIN_NT-10.0 3.2.0(0.340/5/3) 2021-03-29 08:42 x86_64 Cygwin
Cygwin Package Information
Package              Version        Status
mintty               3.5.0-1        OK
====

Regards,

Lem

Re: mintty overstrokes some fonts unexpectedly

[Thomas Wolff]

Am 25.04.2021 um 15:41 schrieb Lemures Lemniscati via Cygwin:
> Hi!
>
> mintty overstrokes some fonts unexpectedly.
> https://gitlab.com/test.cases/mintty-test/-/tree/54ae800e695ecd1741851cab57320a9d0e95a6fd
>
>
> How to reproduce
>
> Setting mintty to use msgothic.ttc in 20pt, run a
> following command.
>
> ====
> perl -e '
> binmode STDOUT, ":utf8";
> print <<END;
> 01234567
> \x{3042}\x{3044}\x{3046}\x{3048}
> AaBbCcDe
> \x{25cb}\x{25cb}\x{25cb}\x{25cb}
> \x{3007}\x{3007}\x{3007}\x{3007}
> \x{300c}\x{300e}\x{3010}\x{3012}
> \x{300d}\x{300f}\x{3011}\x{3013}
> \x{ff62}\x{ff63}\x{ff62}\x{ff63}\x{ff62}\x{ff63}\x{ff62}\x{ff63}
> \x{ff71}\x{ff72}\x{ff73}\x{ff74}\x{ff75}\x{ff76}\x{ff77}\x{ff78}
> END
> '
> ====
>
>
> I got a result mintty-sample-msgothic.png.
> https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/mintty-sample-msgothic.png
>
> In the 4th line of the output, fonts (of u+25cb) were overstruck
> unexpectedly.
> And there are other characters also, which are similarly overstruck.
This is a Windows bug. Mintty clearly instruct Windows to apply 
equidistant spacing to achieve fixed-width character cell behaviour. But 
for certain character ranges, Windows ignores that. Another example for 
such misbehaviour is the Tibetan block (U+0F00-U+0FFF). Mintty could 
work around that by rendering characters separately, at a significant 
penalty for output speed however. Or it could do that only for affected 
ranges, but criteria to identify them are obscure.
Thomas

Re: mintty overstrokes some fonts unexpectedly

[Lemures Lemniscati]

On Sun, 25 Apr 2021 22:33:57 +0200, Thomas Wolff
> 
> Am 25.04.2021 um 15:41 schrieb Lemures Lemniscati via Cygwin:
> > Hi!
> >
> > mintty overstrokes some fonts unexpectedly.
> > https://gitlab.com/test.cases/mintty-test/-/tree/54ae800e695ecd1741851cab57320a9d0e95a6fd
> >
> > I got a result mintty-sample-msgothic.png.
> > https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/mintty-sample-msgothic.png
> >
> > In the 4th line of the output, fonts (of u+25cb) were overstruck
> > unexpectedly.
> > And there are other characters also, which are similarly overstruck.
> This is a Windows bug. Mintty clearly instruct Windows to apply equidistant spacing to achieve fixed-width character cell behaviour. But for certain character ranges, Windows ignores that. Another example for such misbehaviour is the Tibetan block (U+0F00-U+0FFF). Mintty could work around that by rendering characters separately, at a significant penalty for output speed however. Or it could do that only for affected ranges, but criteria to identify them are obscure.

Thank you, Thomas.

I tried some earlier versions of mintty:

* mintty-3.1.0-1 has the same issue
* mintty-2.9.6-0 works expectedly in this case.

Regards,

Lem

Re: mintty overstrokes some fonts unexpectedly

[Thomas Wolff]

Am 26.04.2021 um 01:14 schrieb Lemures Lemniscati via Cygwin:
> On Sun, 25 Apr 2021 22:33:57 +0200, Thomas Wolff
>> Am 25.04.2021 um 15:41 schrieb Lemures Lemniscati via Cygwin:
>>> Hi!
>>>
>>> mintty overstrokes some fonts unexpectedly.
>>> https://gitlab.com/test.cases/mintty-test/-/tree/54ae800e695ecd1741851cab57320a9d0e95a6fd
>>>
>>> I got a result mintty-sample-msgothic.png.
>>> https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/mintty-sample-msgothic.png
>>>
>>> In the 4th line of the output, fonts (of u+25cb) were overstruck
>>> unexpectedly.
>>> And there are other characters also, which are similarly overstruck.
>> This is a Windows bug. Mintty clearly instruct Windows to apply equidistant spacing to achieve fixed-width character cell behaviour. But for certain character ranges, Windows ignores that. Another example for such misbehaviour is the Tibetan block (U+0F00-U+0FFF). Mintty could work around that by rendering characters separately, at a significant penalty for output speed however. Or it could do that only for affected ranges, but criteria to identify them are obscure.
> Thank you, Thomas.
>
> I tried some earlier versions of mintty:
>
> * mintty-3.1.0-1 has the same issue
> * mintty-2.9.6-0 works expectedly in this case.
My previous comment was wrong, sorry. The cause of the issue is that the 
Windows CJK fonts are designed for CJK ambiguous-wide layout. If you run 
mintty with option Charwidth=ambig-wide, or better in an ambiguous-wide 
locale like LC_CTYPE=C.utf8@cjkwide, width handling and font rendering 
will match.
Mintty applies auto-narrowing to some characters that would render far 
out of their character cell, squeezing them into the cell, in order to 
optimise the trade-off between readability and authentic rendering.
Some character ranges were taken out of that mechanism in 2.9.7, 
including the Geometric Shapes which you've encountered, in this case 
because they are "geometric" and the assumption was that they should not 
be tampered for rendering.

Thomas

Re: mintty overstrokes some fonts unexpectedly

[Lemures Lemniscati]

On Mon, 26 Apr 2021 20:31:38 +0200, Thomas Wolff
> 
> Am 26.04.2021 um 01:14 schrieb Lemures Lemniscati via Cygwin:
> > On Sun, 25 Apr 2021 22:33:57 +0200, Thomas Wolff
> >> Am 25.04.2021 um 15:41 schrieb Lemures Lemniscati via Cygwin:
> >>> Hi!
> >>>
> >>> mintty overstrokes some fonts unexpectedly.
> >>> https://gitlab.com/test.cases/mintty-test/-/tree/54ae800e695ecd1741851cab57320a9d0e95a6fd
> >>>
> >>> I got a result mintty-sample-msgothic.png.
> >>> https://gitlab.com/test.cases/mintty-test/-/blob/54ae800e695ecd1741851cab57320a9d0e95a6fd/mintty-sample-msgothic.png
> >>>
> >>> In the 4th line of the output, fonts (of u+25cb) were overstruck
> >>> unexpectedly.
> >>> And there are other characters also, which are similarly overstruck.
> >> This is a Windows bug. Mintty clearly instruct Windows to apply equidistant spacing to achieve fixed-width character cell behaviour. But for certain character ranges, Windows ignores that. Another example for such misbehaviour is the Tibetan block (U+0F00-U+0FFF). Mintty could work around that by rendering characters separately, at a significant penalty for output speed however. Or it could do that only for affected ranges, but criteria to identify them are obscure.
> > Thank you, Thomas.
> >
> > I tried some earlier versions of mintty:
> >
> > * mintty-3.1.0-1 has the same issue
> > * mintty-2.9.6-0 works expectedly in this case.
> My previous comment was wrong, sorry. The cause of the issue is that the Windows CJK fonts are designed for CJK ambiguous-wide layout. If you run mintty with option Charwidth=ambig-wide, or better in an ambiguous-wide locale like LC_CTYPE=C.utf8@cjkwide, width handling and font rendering will match.
> Mintty applies auto-narrowing to some characters that would render far out of their character cell, squeezing them into the cell, in order to optimise the trade-off between readability and authentic rendering.
> Some character ranges were taken out of that mechanism in 2.9.7, including the Geometric Shapes which you've encountered, in this case because they are "geometric" and the assumption was that they should not be tampered for rendering.

Thank you, Thomas!

I overlooked 'man mintty'.

It works expectedly under the setting.


I'd be happy if the option Charwidth is accessible from GUI.

Regards,

Lem