[Python-Dev] 2.1c1 "make test" failures on SGI Irix

[Guido van Rossum]

[Mark Favas]
> Further investigations withe current CVS 2.1c1 (and everything compiled
> without optimization) on a large multi=processor Irix 6.5 box with SGI's
> MIPSpro 7.30 compiler:
> 
> The previously reported failure of test_largefile was due to running
> "make test" on an NFS-mounted filesystem from a box that didn't support
> large files. Works on a local filesystem.
> 
> The previously reported failure of test_zlib looks as though it is due
> to Irix supplying as standard zlib version 1.0.4, whereas the zlib
> module requires at least version 1.1.3. This won't be the only platform
> where an incompatible zlib is system-supplied and built by default. An
> autoconf test for the right version seems indicated (unless setup.py can
> just extract the version string from <wherever>/zlib.h - it's there as
> #define ZLIB_VERSION "1.0.4" on Irix, and #define ZLIB_VERSION "1.1.3"
> on Solaris 8 (both in /usr/include)).

I'll leave that to Andrew, I have no understanding of setup.py.
(Unfortunately Andrew seems out of town. :-( )

> Tests still failing under Irix:
> 
> python Lib/test/test_locale.py
> '%f' % 1024 =? '1,024.000000' ...
> Traceback (most recent call last):
>   File "Lib/test/test_locale.py", line 29, in ?
>     testformat("%f", 1024, grouping=1, output='1,024.000000')
>   File "Lib/test/test_locale.py", line 18, in testformat
>     result = locale.format(formatstr, value, grouping = grouping)
>   File "/var/tmp/mark/src/Lib/locale.py", line 137, in format
>     fields[0],seps=_group(fields[0])
> ValueError: unpack sequence of wrong size

Can you find out what at this point the value of
localeconv()['grouping'] is?  The only way this can happen, I think,
is for that to be a false value -- then _group() returns a single
string value instead of a tuple.  This seems to be a bug in _group()
-- the only place that uses it (format()) always assumes it returns a
tuple of two elements.

I'm not sure what the intention was...

Martin, can you suggest a way out here?

--Guido van Rossum (home page: http://www.python.org/~guido/)