[SciPy-dev] bdist_rpm failure with latest scipy (and f2py) from CVS
Vincent N. Virgilio
virgilio at ieee.org
Mon May 10 10:00:07 EDT 2004
Pearu Peterson wrote:
> On Sun, 9 May 2004, Vincent N. Virgilio wrote:
>
>
>>Then I tried to package scipy. I issued "python2.3 setup.py bdist_rpm",
>>and got a similar failure as in f2py. It seems that atlas_version.so
>>gets built in "site-packages" and not "site-packages/...linalg". RPM
>>complains "installed but not packaged". No amount of fiddling with
>>setup_linalg.py or setup_atlas_version.py could change this. I am in
>>over my head.
>>
>>Have others seen this error? Is there a fix?
>
>
> Try the latest CVS again. Now atlas_version.so is not installed anymore
> while executing bdist_rpm command. Also note that in order to run
> bdist_rpm for scipy, you have to have recent scipy_core installed in your
> system. Otherwise you'll get 'scipy_distutils not found' errors.
> Btw, does anybody know how to set environment for bdist_rpm? If that would
> be possible then one does not have to install scipy_core for creating
> rpm's for scipy.
>
> To summarize the steps needed for creating scipy rpm's and installing
> them:
> 1) Create Scipy_core rpm and install it.
> 2) Create F2py rpm and install it.
> 3) Create Scipy rpm and install it.
>
> Note that it is not possible to create all rpm's once and then installing
> them because latter steps require previous software installed.
>
> Also, before hitting bdist_rpm command, you may need to execute
>
> rm -rf /var/tmp/SciPy-buildroot
>
> when previous bdist_rpm failed.
>
> HTH,
> Pearu
>
> _______________________________________________
> Scipy-dev mailing list
> Scipy-dev at scipy.net
> http://www.scipy.net/mailman/listinfo/scipy-dev
>
Thanks, I'll take this for a spin at the next opportunity.
So apparently it is not quite right to obtain scipy_distutils from f2py?
It seemed to work . . . though I will of course proceed as you suggest.
I was definitely missing the 'rm' step above; I thought it enough to
delete the local 'build' directory.
Thanks again,
Vince Virgilio
More information about the SciPy-Dev
mailing list