[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: test failures on OS X 10.4.11 with Python 2.5.1

From: Phil Schumm <pschumm_at_uchicago.edu>
Date: 2008-01-01 18:39:55 CET

On Dec 29, 2007, at 3:59 PM, Lieven Govaerts wrote:
> The 'parallel' mode was never backported to 1.4.x, so the patch you
> refer to is only applicable to trunk and not related to the issue
> you're seeing.
> I've personally never tried running the tests over python 2.5.x. Do
> you get the same failures with python 2.3.5?

Thanks very much for your reply. And, to answer your question, no --
I don't get these failures with Python 2.3.5 (stock installation on
OS X 10.4). Interestingly, I have now repeated the process several
times, and have observed that the specific tests that fail are a bit
different each time, though they always fail with the same type of

   File "/usr/local/src/subversion-1.4.6/subversion/tests/cmdline/
svntest/main.py", line 286, in run_command_stdin
     pid, wait_code = os.wait()
OSError: [Errno 10] No child processes

Also, I can reduce the number of failures by increasing the load on
the machine (i.e., loading with other concurrent processes, switching
to battery power to reduce the processor speed), and, in some cases,
even eliminate all of the failures entirely (i.e., resulting in all
tests passing). Thus, I don't believe these are real test failures,
but rather an intermittent problem with the testing framework
itself. FWIW, this is a PowerBook G4 (10.4.11), and I'm compiling
with gcc 4.0.1 (latest update of developer tools).

I apologize in advance for following immediately with another
question, but, unfortunately, I am now having problems getting the
Python bindings to compile. As before, this process all worked fine
on the same machine immediately beforehand with Python 2.3.5. With
Python 2.5.1, when I run "make swig-py", the compiler runs for a
while, but then dies with this message:

lipo: can't open input file: /var/tmp//cczAWRg4.out (No such file or
make: *** [subversion/bindings/swig/python/_core.la] Error 1

I checked, and the file subversion/bindings/swig/python/_core.la does
not in fact exist, though I noticed that in my previous Subversion
installation (Python 2.3.5) it did. Also, FWIW, this error was
preceded by several warnings of the following basic types (all just

./subversion/bindings/swig/python/svn_client.c:1193: warning:
initialization discards qualifiers from pointer target type

libtool: link: warning: `/usr/lib/gcc/powerpc-apple-
darwin8/4.0.1/../../..//libiconv.la' seems to be moved

ld: warning multiple definitions of symbol _locale_charset

/usr/libexec/gcc/i686-apple-darwin8/4.0.1/ld: warning ../../../../
libsvn_swig_py-1.dylib cputype (18, architecture ppc) does not match
cputype (7) for specified -arch flag: i386 (file not loaded)

I have no idea whether any of these are expected, or whether they are
related to the terminal error. Any suggestions for how to go about
diagnosing (and hopefully correcting) this problem would be very much

Thanks again,

-- Phil

P.S. In case it's helpful, here's my configure invocation:

   $ ./configure --without-berkeley-db --with-ssl --with-zlib PYTHON=/
work/Versions/2.5/ --with-apxs=/usr/local/apache2/bin/apxs --with-
apr=/usr/local/apache2 --with-apr-
util=/usr/local/apache2 --with-neon=/usr/local --with-swig=/usr/local
-prefix=/usr/local CFLAGS=-I/usr/local/include

The CFLAGS setting is because I have gettext installed on the
machine, and, without it, "make swig-py" complains that it can't find
libintl.h. Alternatively, I have also tried just disabling gettext
support entirely (i.e., with --disable-nls). This seems to have no
effect on the error described above.

To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Jan 1 18:40:21 2008

This is an archived mail posted to the Subversion Users mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.