From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux 2.2.19-6.2.1smp i686; en-US; rv:0.9.1) Gecko/20010607 Description of problem: The math mode sub- and super-script spacing such as that in $\delta_\mathrm{io}$ and $27^{\circ}$ is incorrectly rendered to PostScript by dvips in certain circumstances. These errors do NOT occur with tetex-1.0.6-11, but probably occur with all tetex package relases since Feb 6 2001. How reproducible: Always Steps to Reproduce: I will submit a sample tex file which reproduces the errors (and contains additional detailed information on the bug). Reproduce the error with: 1. latex tetex_rawhide_error.tex 2. dvips -o tetex_rawhide_error.ps tetex_rawhide_error.dvi 3. xdvi tetex_rawhide_error.dvi 4. gv tetex_rawhide_error.ps 5. lpr tetex_rawhide_error.ps (only if you have a PS printer) Actual Results: Look at the math mode sub- and super-script spacing in xdvi compared with the PostScript. The PostScript versions are incorrect in several places in the document. Expected Results: The dvi and PostScript versions should match EXACTLY. Additional info: I am using tetex-1.0.7-22 from RedHat rawhide built from SRPM on a (not terribly customized) RedHat 6.2 machine (rpm --rebuild tetex-1.0.7-22.src.rpm followed by an rpm -Fhv ...). The binary packages are dependent upon a libstdc++ library to which I do not wish to upgrade. I have good reason to believe this problem will be exhibited by the binary packages, also (see below where I fix the problem by editing the SRPM spec file). I have made sure to replace my teTeX configuration files (texmf.cnf and mktex.cnf) with the .rpmnew ones generated from the package install. I have recomplied the rawhide SRPM after editing the spec file to exclude the ASCII Corp. dvips patches for Japanese support which was enabled in the spec file on Feb 6 2001 for bug 26249 (details of my spec file are included as an attachment in my bugzilla report), and the resulting dvips works without error. Therefore, I do not think the existing packaging of teTeX should continue to have Japanese support until these patches do not break dvips. This is a SERIOUS ERROR which discredits the integrity of the mature teTeX distribution and which will surely cause users to produce incorrect PostScript output. I view such an error as serious as "loss of data", so I feel like this bug should be classified as more than "normal" severity (esp. since the bug is entirely due to enabling Japanese support to fix another bug).
Created attachment 22287 [details] LaTeX file to recreate the bug (also contains additional details)
Created attachment 22288 [details] a diff of my (hacked) spec file which allows the package to be built without Japanese support so that the errors are not created
Yes, this happens with the precompiled binary packages too. Turning off Japanese support entirely is a little drastic, and 'Japanese doesn't work at all' is more severe than this bug. Is the problem definitely in the dvipsk patch, I wonder?
The problem is solely due to the patch for dvips in dvipsk-jpatch-p1.5d.tar.gz. I verified this in the tetex-1.0.7-24 SRPM by commenting out the following line (line 292) of the spec file: patch -d texk -p0 < dvipsk-j/dvipsk586.patch As packaged, tetex-1.0.7-24 has the dvips error I have described, but when this single line is commented out, the error goes away (and luckily none of the other patches fail!). I can see your position in needing to have support for Kanji in your teTeX distribution for all the Japanese RedHat users, but I'm disappointed that yet again a large software vendor is choosing to add features immediately and fix (known) bugs later. I suppose I'll have to roll my own teTeX from now on.
I forwarded this bug detail to ASCII Corp. They have verified the bug and provided a patch (see the URL below for the patch---notice that it will not be available until tomorrow). It looks like this patch should make its way into rawhide tetex (but as I haven't seen the patch, I have not tried it yet). I have attached a mail I received from one of the ASCII Corp. developers. Please note that they recommend that their patched version of dvips should NOT be provided on the default version of any teTeX distribution. Furthermore, note that on ASCII Corp's advice SuSE provides standard dvips as is and pdvips for the patched Japanese one. I really think this is what RedHat should do (but maybe with a more sensible name like dvipsj). Regards, John ================================================== From hideak-t.jp Wed Jul 25 10:39:52 2001 Date: Wed, 25 Jul 2001 16:03:06 +0900 From: Hideaki Togashi <hideak-t.jp> To: jhs28.ac.uk Cc: www-ptex.jp, hideak-t.jp Subject: Re: [www-ptex 1946] Re: dvipsk-jpatch-p1.5d wysiwyg error in mathmode Hello, Mr. John H.Steele. This is Hideaki Togashi of ASCII Corp. in Japan. At last we may be able to specify the problem, we think. Your ps files and our Japanese TeX developers help us greatly. Thank you very much for your help. Now we made the new Japanese patch to dvipsk. For the time being, we offer the new patch as beta version at the following URL; ftp://ftp.ascii.co.jp/pub/TeX/ascii-ptex/beta/dvipsk-jpatch-p1.5e.tar.gz This file will be placed tomorrow. If you check this new patch and report the result to us, we will thank you greatly again. And if nobody report the bug, we plan to make this file open as the official version of the new Japanese patch to dvipsk. Thank you again. By the way, is the Japanese patched dvips used as the only one dvips on the RedHat Linux? We do not recommend it. When the SuSE Linux cheam asked us, we answered the same. So on the international version of SuSE Linux 7.2, Japanese patched dvips is named 'pdvips', and 'dvips' is the original one. We cannot check all the environment, and like you this time, some incompatibilities are reported sometimes. When we get the reports we try to do our best, but we cannot declare that Japanese patched dvips is full compatible to the original one. If you can, please tell the RedHat cheam not to use the japanese patched dvips as the only one dvips. Regards, -- Hideaki Togashi <hideak-t.jp>
Thank you very much for this feeback, and sorry for the delay in responding. I am about to build with this patch, and I will look at shipping two dvips programs.
(The patch fixes the problem here.)
Fixed in 1.0.7-27.