Intallation Tests Failed
I just finished downloading and installing the latest NumPy and SciPy; the installation tests failed: Ran 1552 tests in 12.197s FAILED (errors=196) I followed the instructions on the wiki installation page: first NumPy, then BLAS (as a directory within /usr/local/scipy-0.5.0), then LAPACK (also within that directory), then SciPy. When compiling LAPACK I reran it without the quotation marks for the optimization option on the command line; hadn't seen options quoted before. Could that make a difference? Also, when LAPACK finished compiling it did not automatically return control to the command line; I had to kill the process with ctrl-c. What do you suggest I do to get a working installation? I can send the test results if that helps. Thanks, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
Rich Shepard wrote:
I just finished downloading and installing the latest NumPy and SciPy; the installation tests failed:
Ran 1552 tests in 12.197s
FAILED (errors=196)
I followed the instructions on the wiki installation page: first NumPy, then BLAS (as a directory within /usr/local/scipy-0.5.0), then LAPACK (also within that directory), then SciPy.
When compiling LAPACK I reran it without the quotation marks for the optimization option on the command line; hadn't seen options quoted before. Could that make a difference? Also, when LAPACK finished compiling it did not automatically return control to the command line; I had to kill the process with ctrl-c.
What do you suggest I do to get a working installation? I can send the test results if that helps.
What is your platform? What does numpy.test() do by itself? What are the failing tests (most likely they are all related to a few points of failure so you probably don't have to provide all the failing tests. Have you tried with a pre-built ATLAS? -Travis
On Mon, 28 Aug 2006, Travis Oliphant wrote:
What is your platform? What does numpy.test() do by itself?
My apologies, Travis. After sending the message I realized that I should have specified I'm running Slackware-10.2. I'm rebuilding numpy right now; the directory is /usr/local/numpy-1.0b4.
What are the failing tests (most likely they are all related to a few points of failure so you probably don't have to provide all the failing tests.
Let me rebuild everything (using OPT = "-O2" rather than OPT = -O2 for lapack) and see if it's strickly a user error here. Some of the failed tests are: ERROR: binary hit-or-miss transform (1-3) ERROR: iterating a structure (1-3) ERROR: label (1-13) ERROR: maximum (1, 2) ERROR: maximum filter (6-9) ERROR: maximum position (1, 3) ERROR: mean (1, 2) ERROR: minimum (1, 2) ERROR: minimum filter (6-9) and others in morphological, rank, median, standard deviation, sum, variance, watershed_ift, white tophat, and check_rvs.
Have you tried with a pre-built ATLAS?
No. That was my next step. While running the numpy installation, 'python setup.py install >& install.log & tail -f install log,' I guess it finished at the line: copying build/src.linux-i686-2.4/numpy/core/ufunc_api.txt -> /usr/lib/python2.4/site-packages/numpy/core/include/numpy because the tail sat there for more than 10 minutes. I've attached the gzipped log file. I don't see the syntax to test only numpy. Thanks, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Travis Oliphant wrote:
What does numpy.test() do by itself?
Travis, Ran 481 tests in 2.744s OK So, on to the next step. I'm building Atlas as I write. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Travis Oliphant wrote:
What does numpy.test() do by itself?
Travis, I may have found part of the problem. Having rebuilt and re-installed in this order: atlas, blas, lapack, numpy, numpy.test() now reports 481 tests successfully completed. On to scipy rebuild/re-installation. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Travis Oliphant wrote:
What are the failing tests (most likely they are all related to a few points of failure so you probably don't have to provide all the failing tests.
SciPy rebuilt. Invoke ipython. Enter 'import numpy;' prompt for next line returned. Enter 'import scipy' and the following is returned: In [2]: import scipy Overwriting info=<function info at 0x40a97c34> from scipy.misc.helpmod (was <function info at 0x408f5bc4> from numpy.lib.utils) Overwriting who=<function who at 0x40a97aac> from scipy.misc.common (was <function who at 0x408f5b1c> from numpy.lib.utils) Overwriting source=<function source at 0x40a97c6c> from scipy.misc.helpmod (was <function source at 0x408f5bfc> from numpy.lib.utils) If instead I type from numpy import * from scipy import * I see: In [2]: from scipy import * Overwriting info=<function info at 0x40a97c34> from scipy.misc.helpmod (was <function info at 0x408f5bc4> from numpy.lib.utils) Overwriting who=<function who at 0x40a97aac> from scipy.misc.common (was <function who at 0x408f5b1c> from numpy.lib.utils) Overwriting source=<function source at 0x40a97c6c> from scipy.misc.helpmod (was <function source at 0x408f5bfc> from numpy.lib.utils) /usr/lib/python2.4/site-packages/numpy/dft/__init__.py:2: UserWarning: The dft subpackage will be removed by 1.0 final -- it is now called fft warnings.warn("The dft subpackage will be removed by 1.0 final -- it is now called fft") Running 'test(level=1, verbosity=2)' there are still failures: Ran 1552 tests in 11.576s FAILED (errors=196) What do you suggest? Thanks, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
Rich Shepard wrote:
I just finished downloading and installing the latest NumPy and SciPy; the installation tests failed:
Wait a minute. By the "latest" SciPy what do you mean?? The SciPy release is behind the NumPy release. Only the SVN version of SciPy works with the latest NumPy. I suspect the errors are due to that. -Travis
On Mon, 28 Aug 2006, Travis Oliphant wrote:
The SciPy release is behind the NumPy release. Only the SVN version of SciPy works with the latest NumPy.
I suspect the errors are due to that.
Travis, Aha! I downloaded numpy-1.0b4 and scipy-0.5.0 from the SciPy site. So, I'll grab the SVN version of SciPy and try again. Do I also need the SVN version of numpy? Oh, what the heck, I'll redo both from the trunk. Thanks, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
Rich Shepard wrote:
On Mon, 28 Aug 2006, Travis Oliphant wrote:
The SciPy release is behind the NumPy release. Only the SVN version of SciPy works with the latest NumPy.
I suspect the errors are due to that.
Travis,
Aha! I downloaded numpy-1.0b4 and scipy-0.5.0 from the SciPy site. So, I'll grab the SVN version of SciPy and try again. Do I also need the SVN version of numpy? Oh, what the heck, I'll redo both from the trunk.
It's easy enough if you are building from sources to just grab the SVN of both. My focus is to keep both synchronized so they work together. Yes, you will need the trunk from both at this point. Sorry, I didn't catch that earlier. -Travis
On Mon, 28 Aug 2006, Travis Oliphant wrote:
It's easy enough if you are building from sources to just grab the SVN of both. My focus is to keep both synchronized so they work together.
Yes, you will need the trunk from both at this point. Sorry, I didn't catch that earlier.
Travis, I should have remembered reading that, too. After dinner I'll start over again with both. Not many files were changed with the numpy checkout, but scipy revision 2180 brought down many files. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Rich Shepard wrote:
After dinner I'll start over again with both.
Well, I decided to do it before dinner.
Not many files were changed with the numpy checkout, but scipy revision 2180 brought down many files.
Closer. Many warnings scrolled by during the build/install. After invoking ipython, 'import scipy' showed: In [2]: import scipy import misc -> failed: cannot import name place Ran 'scipy.test(level=1)' and I'm down from 196 errors to 1: ====================================================================== ERROR: check_integer (scipy.io.tests.test_array_import.test_read_array) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/lib/python2.4/site-packages/scipy/io/tests/test_array_import.py", line 55, in check_integer from scipy import stats File "/usr/lib/python2.4/site-packages/scipy/stats/__init__.py", line 7, in ? from stats import * File "/usr/lib/python2.4/site-packages/scipy/stats/stats.py", line 1693, in ? import distributions File "/usr/lib/python2.4/site-packages/scipy/stats/distributions.py", line 16, in ? from numpy import atleast_1d, polyval, angle, ceil, place, extract, \ ImportError: cannot import name place ---------------------------------------------------------------------- Ran 1388 tests in 11.423s FAILED (errors=1) Out[4]: <unittest.TextTestRunner object at 0x41a4144c> Also, **************************************************************** WARNING: clapack module is empty ----------- See scipy/INSTALL.txt for troubleshooting. Notes: * If atlas library is not found by numpy/distutils/system_info.py, then scipy uses flapack instead of clapack. **************************************************************** ATLAS _is_ installed; perhaps in the wrong place? It was built within /usr/local/ATLAS, not as a scipy/ subdirectory. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Rich Shepard wrote:
Not many files were changed with the numpy checkout,
On my notebook I removed the numpy directory and grabbed the trunk head from the svn server. After that built, I ran 'numpy.test(level=1)' and got three errors: check_ascii, check_both, and check_place. ====================================================================== ERROR: check_ascii (numpy.core.tests.test_multiarray.test_fromstring) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/core/tests/test_multiarray.py", line 120, in check_ascii a = fromstring('1 , 2 , 3 , 4',sep=',') ValueError: don't know how to read character strings for given array type ====================================================================== ERROR: check_both (numpy.lib.tests.test_function_base.test_extins) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/lib/tests/test_function_base.py", line 249, in check_both place(a,mask,0) NameError: global name 'place' is not defined ====================================================================== ERROR: check_place (numpy.lib.tests.test_function_base.test_extins) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/lib/tests/test_function_base.py", line 242, in check_place place(a,[0,1,0,1,0,1,0],[2,4,6]) NameError: global name 'place' is not defined ---------------------------------------------------------------------- Ran 481 tests in 0.863s FAILED (errors=3) Out[5]: <unittest.TextTestRunner object at 0x409b632c> Something I did incorrectly? Thanks, Travis, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
Rich Shepard wrote:
On Mon, 28 Aug 2006, Rich Shepard wrote:
Not many files were changed with the numpy checkout,
On my notebook I removed the numpy directory and grabbed the trunk head from the svn server. After that built, I ran 'numpy.test(level=1)' and got three errors: check_ascii, check_both, and check_place.
====================================================================== ERROR: check_ascii (numpy.core.tests.test_multiarray.test_fromstring) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/core/tests/test_multiarray.py", line 120, in check_ascii a = fromstring('1 , 2 , 3 , 4',sep=',') ValueError: don't know how to read character strings for given array type
This error is known on Python 2.3. I wouldn't worry about it. It's probably a missing function that Python added to it's C-API in Python 2.4
====================================================================== ERROR: check_both (numpy.lib.tests.test_function_base.test_extins) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/lib/tests/test_function_base.py", line 249, in check_both place(a,mask,0) NameError: global name 'place' is not defined
====================================================================== ERROR: check_place (numpy.lib.tests.test_function_base.test_extins) ---------------------------------------------------------------------- Traceback (most recent call last): File "/usr/local/lib/python2.3/site-packages/numpy/lib/tests/test_function_base.py", line 242, in check_place place(a,[0,1,0,1,0,1,0],[2,4,6]) NameError: global name 'place' is not defined
Both of these errors indicate some problem with the installation. Are you sure you have the right SVN tree? What is the output of svn info (run from the same directory you ran setup.py from) What about python -c "import numpy; print numpy.__version__"
On Mon, 28 Aug 2006, Travis Oliphant wrote:
This error is known on Python 2.3. I wouldn't worry about it. It's probably a missing function that Python added to it's C-API in Python 2.4
[rshepard@salmo ~]$ ls /var/log/packages/ | grep python python-2.4.1-i486-1 Apparently not.
Both of these errors indicate some problem with the installation. Are you sure you have the right SVN tree? What is the output of svn info (run from the same directory you ran setup.py from)
Workstation: [rshepard@salmo /usr/local/scipy]$ svn info Path: . URL: http://svn.scipy.org/svn/scipy/trunk Repository Root: http://svn.scipy.org/svn/scipy Repository UUID: d6536bca-fef9-0310-8506-e4c0a848fbcf Revision: 2180 Node Kind: directory Schedule: normal Last Changed Author: oliphant Last Changed Rev: 2180 Last Changed Date: 2006-08-27 00:44:15 -0700 (Sun, 27 Aug 2006) Properties Last Updated: 2006-08-28 17:38:34 -0700 (Mon, 28 Aug 2006)
What about python -c "import numpy; print numpy.__version__"
[rshepard@salmo ~]$ python -c "import numpy; print numpy.__version__" 1.0b4 Notebook: Path: . URL: http://svn.scipy.org/svn/numpy/trunk Repository Root: http://svn.scipy.org/svn/numpy Repository UUID: 94b884b6-d6fd-0310-90d3-974f1d3f35e1 Revision: 3089 Node Kind: directory Schedule: normal Last Changed Author: oliphant Last Changed Rev: 3089 Last Changed Date: 2006-08-28 13:56:23 -0700 (Mon, 28 Aug 2006) Properties Last Updated: 2006-08-28 17:52:36 -0700 (Mon, 28 Aug 2006) Path: . URL: http://svn.scipy.org/svn/scipy/trunk Repository Root: http://svn.scipy.org/svn/scipy Repository UUID: d6536bca-fef9-0310-8506-e4c0a848fbcf Revision: 2180 Node Kind: directory Schedule: normal Last Changed Author: oliphant Last Changed Rev: 2180 Last Changed Date: 2006-08-27 00:44:15 -0700 (Sun, 27 Aug 2006) Properties Last Updated: 2006-08-28 17:47:40 -0700 (Mon, 28 Aug 2006) Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
On Mon, 28 Aug 2006, Travis Oliphant wrote:
Both of these errors indicate some problem with the installation. Are you sure you have the right SVN tree? What is the output of
Eureka! I checked out the numpy svn trunk on my workstation and built/installed that. Now scipy generates warnings on the tests, but no errors. Ran 1569 tests in 3.977s OK Whew! I'll see what's different on the notebook later; now I need to get some productive work done on my application. Thanks, Travis, Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
Rich Shepard wrote:
On Mon, 28 Aug 2006, Travis Oliphant wrote:
Both of these errors indicate some problem with the installation. Are you sure you have the right SVN tree? What is the output of
Eureka!
I checked out the numpy svn trunk on my workstation and built/installed that. Now scipy generates warnings on the tests, but no errors.
Printed warnings are no big deal. Only test failures and errors are signficant. Congradualations. I suspect the big problem was the confusion over which SciPy works with which NumPy? We need to make that clear. Hopefully by the first of next week, the two will be in sync and stay that way for a long time (in that SciPy will work with all Numpy 1.0.X releases. That is the goal. -Travis
On Tue, 29 Aug 2006, Travis Oliphant wrote:
Printed warnings are no big deal. Only test failures and errors are signficant.
Travis, That was my assumption. Warnings can be cleaned up later without affecting usability.
Congradualations. I suspect the big problem was the confusion over which SciPy works with which NumPy? We need to make that clear. Hopefully by the first of next week, the two will be in sync and stay that way for a long time (in that SciPy will work with all Numpy 1.0.X releases.
I'd like to suggest that the version numbers then be synchronized. It's not obvious to folks like me that numpy-1.0b4 and scipy-0.5.0 might (or might not) work together. If the numbers are identical it's a good bet us novices will figure out how to make a matched pair. The next issue I need to test and, I hope, not need to resolve is the svn trunk checkout of matplotlib to go with numpy and scipy. Whew! At least I have my pysqlite issues fixed up now. Progress on several fronts today. Rich -- Richard B. Shepard, Ph.D. | The Environmental Permitting Applied Ecosystem Services, Inc.(TM) | Accelerator <http://www.appl-ecosys.com> Voice: 503-667-4517 Fax: 503-667-8863
participants (3)
-
Rich Shepard -
Travis Oliphant -
Travis Oliphant