# A difficult CJK Problem - inconsistent Japanese character kerning [closed]

Typesetting the Japanese translation of my free physics book, I am encountering a problem from hell'. When a specific line of Japanese text is added or commented out, the typesetting of a chapter start later on (after a \newpage!) differs. The chapter in question starts about ten pages later, and the difference is the following: the distances between the Japanese characters change.

Due to the complexity, I cannot post a minimal example. The change happens also before the new chapter, but it remains still bizarre.

It seems that some formatting commands in my cls file inadvertently change the kerning of the Japanese characters. Could somebody help here? Is there a way to "reset" the kerning of kanjis and other Japanese characters to a default value? Which command does this?

In am using pdflatex, then dvips, then Distiller. OSX, TexLive 2017. Thank you very much in advance for any help.

OK, I reduced the tracingall logs from 420 to 20 MB. There are two differences: glues values are different, and the kerning values are different. Examples are the very first one

< @\kern via @@1 b=81 p=0 d=8281
< @@3: line 2.1 t=9181 -> @@1
---
> @\kern via @@1 b=90 p=0 d=10000
> @@3: line 2.1 t=10900 -> @@1


And later one, one picked at random

< @\kern via @@65 b=0 p=0 d=100
< @\kern via @@66 b=0 p=0 d=100
< @\kern via @@67 b=9 p=0 d=361
< @\kern via @@68 b=9 p=0 d=361
< @\kern via @@69 b=50 p=0 d=3600
< @\kern via @@70 b=50 p=0 d=3600
< @@91: line 7.1 t=5086 -> @@70
< @@92: line 7.2 t=2013 -> @@66
---
> @\kern via @@62 b=1 p=0 d=121
> @\kern via @@63 b=1 p=0 d=121
> @\kern via @@64 b=13 p=0 d=529
> @\kern via @@65 b=13 p=0 d=529
> @\kern via @@66 b=63 p=0 d=5329
> @\kern via @@67 b=63 p=0 d=5329
> @@87: line 7.1 t=2388 -> @@65
> @@88: line 7.2 t=2364 -> @@63


There are dozens of such differences between the two log files.

I now used \tracingmacros=1 . The difference between the two texlogs ("short line" / "long line") is found here:

www.motionmountain.net/sl-ll.diff (only 300 lines)

and the two texlogs

www.motionmountain.net/motionmountain-J-volume1-sl2.texlog (10 MB) www.motionmountain.net/motionmountain-J-volume1-ll2.texlog (10 MB)

## closed as unclear what you're asking by Troy, CarLaTeX, user36296, Bobyandbob, Henri MenkeDec 29 '17 at 21:06

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

• you need to give at least some clues, are you using xetex or luatex, or if you are using pdftex which packages are you using to define the Japanese font setup. To track it down, put \tracingall` just before the point where things change and then compare the logs with and without your line commenet out – David Carlisle Dec 6 '17 at 16:09
• pdflatex then dvips? possible but slightly odd, do you mean latex then dvips? (It's almost certainly easier to use xelatex or lualatex for fonts with CJK font requirements these days) – David Carlisle Dec 6 '17 at 17:07
• Yes, pdflatex than dvips, because I have exclusively eps figures. – Motion Mountain Dec 6 '17 at 17:16
• But it is just odd to use pdflatex and set pdfoutput to 0 rather than use latex (which is the same thing) also of course you can include eps into pdlatex (latex will use epstopdf to convert them automatically) – David Carlisle Dec 6 '17 at 17:19
• hmm perhaps \tracingmacros=1 might have been better than \tracingall, the log you show is too late it is just showing that the line breaking is different, it is bound to be different of the characters are differently spaced but you need to see tracing of the macros adding the space – David Carlisle Dec 6 '17 at 19:35