astropy/io/votable/tests/vo_test.py failure on MIPS

[mdboom]

Sorry for bombarding you with exotic failures :-)
In MIPS (little endian), the compilation of version 0.2.1 fails with

________________________________ test_validate _________________________________

    def test_validate():
[...]
        result = validate(get_pkg_data_filename('data/regression.xml'),
>                         output, xmllint=False)

astropy/io/votable/tests/vo_test.py:713: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

source = '/tmp/astropy-test-fOeQY2/lib.linux-mips-2.7/astropy/io/votable/tests/data/regression.xml'
output = <_io.StringIO object at 0x226f170>, xmllint = False
filename = '/tmp/astropy-test-fOeQY2/lib.linux-mips-2.7/astropy/io/votable/tests/data/regression.xml'

    def validate(source, output=sys.stdout, xmllint=False, filename=None):
[...]
        if len(lines):
            xml_lines = iterparser.xml_readlines(content_buffer)
            for warning in lines:
                w = exceptions.parse_vowarning(warning)
                if not w['is_something']:
>                   output.write(warning)
E                   TypeError: unicode argument expected, got 'str'

astropy/io/votable/table.py:242: TypeError

See https://buildd.debian.org/status/fetch.php?pkg=python-astropy&arch=mipsel&ver=0.2.1-1%7Eexp4&stamp=1366656559 for the full log.

[eteq]

This is weird. @olebole - do you know if the python here has any unusual unicode-related features or compilation options?

[olebole]

The python2.7 build log has

cd /build/buildd-python2.7_2.7.4-2-mipsel-7K7PPG/python2.7-2.7.4/build-static && \
      CC="mipsel-linux-gnu-gcc" CFLAGS="-D_FORTIFY_SOURCE=2 -g -Wformat -Werror=format-security " LDFLAGS="-Wl,-z,relro" \
         \
        ../configure \
        --prefix=/usr --enable-ipv6 --enable-unicode=ucs4 --with-dbmliborder=bdb:gdbm --with-system-expat --with-system-ffi --with-fpectl

but this is similar for all platforms. I couldn't find any other unicode specifics.

[mdboom]

The attached I think resolves this problem. Unfortunately, there's another hiding behind that (#1023).

[eteq]

Huh, tricky. So you tested this on a MIPS VM, is that right @mdboom ? (If so, it seems fine to merge.)

[mdboom]

I tested this in a Debian Squeeze installation (python2.6, numpy1.7.1) on qemu-system-mipsel.

[olebole]

With the patches, I can build the package on MIPSel on the current Debian (experimental).