xdvipdfmx-20200116 : additional q/Q bracket around BT...ET

[Hironobu Yamashita]

Hi,

> > > Anyway you have already reverted my change thus I guess you are
> > > going to commit your change.
> >
> > Not really; the reason why I reverted your change on 2019-06-30 is
> > that adding q-Q bracket would break many existing TeX files in which
> > \special commands had been carefully constructed to work fine with
> > TL2019 xdvipdfmx.  I just didn't like such a change in xdvipdfmx
> > between TL2019 and TL2020.
>
> Currently known problem is only clipping and I don't understand how
> so many TeX files could reply on such feature.

I don't know either, but personally this thread was not the first
time for me to receive questions from users of text clipping.

> > Therefore, I don't want to re-commit your previous change of adding
> > q-Q bracket.  On the other hand, I'd like to follow your decision
> > on whether you accept my "0 Tr" patch or not.
>
> I don't understand why you are going to follow my decision for the
> patch prepared by yourself while the decision to reject my previous
> change was made by yourself alone.

Sorry for reverting changes without contacting to you, but please
also note that it was me who transferred the fake-bold issue to you
in https://tug.org/pipermail/dvipdfmx/2019-June/000014.html.
At that time, it *should have been possible* for me to withstand the
fake-bold issue of xdvipdfmx and think of another way to exploit
\special commands to get correct results.  But, I didn't do so.
Now, we have received some questions from users of text clipping.
This made me realized that any request of changing internal behavior
of xdvipdfmx could affect unrelated users.  I should have been more
careful before asking, and sorry for taking your time.

> If you just want suggestions on the correctness of the patch: There
> are two kind of text. Normal and embolden. For normal text, current
> mode setting should be used. For embolded text, mode "2" is used.
> This patch introduces another bug that modes 1, 2, and 3 don't work
> too when embolen text is used before.
>
> After incorporating the provided patch, even single embolden letter
> newly inserted in the footnote of page 16 would cause a drastic
> change in the rendering of page 345 where Tr is used.

Thanks for your kind suggestions.  I'd like to withdraw my patch.


Regards,
Hironobu