Re: bug#6327: sort fails on some UTF-8 input

[Eric Blake]

[redirecting to bug-gnulib]

On 06/02/2010 01:37 PM, Paul Eggert wrote:
> On 06/01/2010 09:51 PM, River Tarnell wrote:
>> I'm using coreutils 8.5 on Solaris 10.
>>
>> GNU 'sort' fails to sort some input, while Solaris 'sort' handles it
>> correctly:
> 
> Amusingly enough, on that same test case I found the same problem
> with GNU 'sort' that you did, but I also found that Solaris 'sort'
> reports that it runs out of memory, even in 64-bit mode.  For example:
> 
> 1010-kiwi $ LC_ALL=en_CA.UTF-8 /usr/bin/sparcv9/sort sort_test.txt 
> sort: insufficient memory; use -S option to increase allocation
> 1011-kiwi $ LC_ALL=en_CA.UTF-8 coreutils-8.5/src/sort sort_test.txt
> coreutils-8.5/src/sort: string comparison failed: Illegal byte sequence
> coreutils-8.5/src/sort: Set LC_ALL='C' to work around the problem.
> coreutils-8.5/src/sort: The strings compared were 
> `\360\222\203\276\360\222\205\226' and 
> `\360\222\200\255\360\222\213\253\360\222\213\253\360\222\200\255'.
> 
> I expect that the exact failure mode probably depends on the
> locale (and/or whether you're using x86 or sparc),
> and that GNU 'sort' checks for strcoll failures but
> Solaris 'sort' does not (and thus crashes).  If my guess is right,
> this appears to be a bug in the Solaris strcoll implementation.
> I don't see a simple workaround.  You might file a bug report
> with Sun.

And in the meantime, now that we've confirmed that it is a Solaris
strcoll() bug, it would be nice to code a gnulib workaround.

-- 
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature