[tex-live] Win32: xelatex -0.999.4 failure to process sample2e.tex

Jonathan Kew jonathan at jfkew.plus.com
Thu Aug 7 12:29:50 CEST 2008


On 7 Aug 2008, at 8:22 AM, Christian Schenk wrote:

> Akira Kakuto wrote:
>> Hi George,
>>> After this morning's update:
>>>
>>> F:\TeXLive\2008\bin\win32>ls -l xe*.*
>>> -rwxrwxrwx  1 WhiteG 0    3072 2008-08-05 05:05 xelatex.exe
>>> -rwxrwxrwx  1 WhiteG 0 2318336 2008-08-05 05:05 xetex.exe
>>>
>>> xetex is working on my real Windows XP system.   The binaries
>>> haven't changed, but the font cache was rebuilt,so the problem
>>> may lie outside xetex.
>> Fortunately I've reproduced your previous crash by compiling
>> on vista. I've identified that the access violation is occuring at
>> applytfmfontmapping().  Please give me some time.
>> I believe it will become more stable later.
>
> MiKTeX-XeTeX had the same issue: a crash in applytfmfontmapping. See  
> the "xelatex math problem with sqrt" thread (XeTeX mailing list).
>
> The MiKTeX bug report:
> https://sourceforge.net/tracker/?func=detail&atid=110783&aid=1977813&group_id=10783

I remember this thread; didn't realize we might be seeing a similar  
issue with Akira's binaries too. Did you ever determine how the  
mapping could be left uninitialized? I thought I'd checked the code  
and it was supposed to zero the appropriate entry in the array, if no  
mapping is actually loaded (and therefore zeroing the entire array  
ahead of time shouldn't be needed).

If it's possible to load a tfm without correctly setting (or zeroing)  
the font_mapping array entry, that would be a bug that could  
potentially affect any platform, and we should fix it ASAP. But when I  
reviewed this previously, I couldn't see how that would happen. Any  
further details would be very welcome.

Thanks,

Jonathan



More information about the tex-live mailing list