[dvipdfmx] forward references

Shunsaku Hirata shunsaku.hirata74 at gmail.com
Tue Jul 21 21:05:06 CEST 2020


This issue should have been resolved by the changes below:

http://www.tug.org/svn/texlive?view=revision&revision=55887

I don't know when it will be available on your platform but please try it
when available.

Dvipdfmx is extended to allow yet-to-be-defined graphics objects to be
used but it can easily result in broken PDFs if they are finally not
defined. A warning is issued in such a case, but since XeTeX invokes
dvipdfmx with "quiet" mode by default users may not notice unless
dvipdfmx is separately invoked.

Thanks,
Shunsaku Hirata

2020年7月15日(水) 7:31 Sašo Živanović <saso.zivanovic at guest.arnes.si>:
>
> Many thanks, just the answer I hoped for. Keeping my fingers crossed,
> Sašo
>
> Shunsaku Hirata je 14. 07. 20 ob 23:48 napisal:
> > Hi,
> >
> >> I have tracked down the problem to what I would naively call the
> >> resolution of "forward references" in xdvipdfmx. "Naively", because I
> >> have no idea how that program works, and I'm not even sure whether to
> >> call the problem a feature or a bug. The issue is explained in detail in
> >> https://github.com/sasozivanovic/memoize/issues/1.
> >
> > As you noted in the github issue above, the problem is that you are trying
> > to reference an object before it is defined. It is possible for dvipdfmx to
> > workaround this problem, however, the problem itself is improper use of
> > object reference and hence not a bug of dvipdfmx.
> >
> > If this issue is simply a problem of "forward references" of objects, it could
> > be resolved relatively easily. However, in this example, a graphics object
> > ("pgfshade1" for shading) is placed before it is defined. This makes it
> > difficult to resolve the issue.
> >
> >> I would appreciate it if you could take the time to see if the second
> >> approach to the resolution of the issue (as proposed in the github
> >> issue) is feasible. And if it is, I would of course be very happy if the
> >> issue could be addressed.
> >
> > I think it is feasible, but it may take some time to inspect the impact of
> > changing dvipdfmx behavior.
> >
> > I will consider how this issue can be resolved.
> >
> >
> > Thanks,
> > Shunsaku Hirata
> >



More information about the dvipdfmx mailing list.