Hi
Sorry for the delay. I will test and patch it tonight.

On 2/21/07, Frank Küster <[EMAIL PROTECTED]> wrote:
# I think dvipdfmx should not be released non-working in some parts of
# the world
severity 403323 serious
tags 403323 patch
thanks

Jin-Hwan Cho <[EMAIL PROTECTED]> wrote:

> On Feb 15, 2007, at 10:43 PM, Frank Küster wrote:
>
>> Dear dvipdfmx developers, dear dvipdfmx maintainer in Debian,
>>
>> in the last weeks, we found that dvipdfm(x) would segfault when
>> launched
>> in timezones with half-hour offset to UTC, and Mark Wicks wrote a
>> patch.  It would be nice if this could be applied to dvipdfmx, too.
>
> The topic had been discussed since Sep. 2005. The history of the patch
> (for dvidpfmx not dvipdfm) can be found at
>
> http://cvs.ktug.or.kr/viewcvs/dvipdfmx/src/pdfdoc.c?view=log
>
> But the CVS tag was not applied until the version 20061211. Recently
> (Dec 2006),
> while discussing TeXLive 2007, the file was updated again.

Ah, that's good.  So TeXlive has a corrected version.  Hm, I looked at
the code and compared

http://cvs.ktug.or.kr/viewcvs/dvipdfmx/src/pdfdoc.c?r1=1.42&r2=1.45

to

http://svn.debian.org/wsvn/debian-tex/texlive/trunk/dpatch/texlive-bin/61_dvipdfm_timezone.dpatch?op=file&rev=0&sc=0

The differences are:

- different hierarchies of ifdef's; in dvipdfm, bd_time->tm_gmtoff takes
  precedence over timezone, in dvipdfmx it's the other way round.  Maybe
  not very important

- In the sprintf format string, you use %03ld and %02ld, which seems
  correct, whereas dvipdfm misses the 'l' length modifier.  Hm, could
  this lead to errors?  It seems to work...

- To me, the computation of the timezone information in date_string
  looks wrong in dvipdfmx.  Ah, no, it is correct.  But

  (tz_offset/60) % 60

  is easier to understand than

  (timezone % 3600) % 60

> I hope the maintainer in Debian would use the recent CVS code.

Umh, no.  Debian is in a freeze, we won't take new upstream versions at
this time.

I think the Yu Guanghui, the Debian maintainer for dvipdfmx, should take
the patch used in dvipdfmx (this is the context diff format:

http://cvs.ktug.or.kr/viewcvs/dvipdfmx/src/pdfdoc.c?r1=1.42&r2=1.45&diff_format=c

) and upload targetted at etch.  Or do you want me to do an NMU?

Regards, Frank
--
Dr. Frank Küster
Single Molecule Spectroscopy, Protein Folding @ Inst. f. Biochemie, Univ. Zürich
Debian Developer (teTeX/TeXLive)



Reply via email to