> On Feb 10, 2016, at 15:55, Derek Homeier 
> <de...@astro.physik.uni-goettingen.de> wrote:
> 
> 
>> On 11 Feb 2016, at 12:35 am, Alexander Hansen <alexanderk.han...@gmail.com> 
>> wrote:
>> 
>>> when building the updated matplotlib package in maintainer mode, after the 
>>> tests are run I am
>>> still getting the dreaded verification error for the .deb:
>>> 
>>> Error: Compiled python module points to fink build dir.
>>> 
>>> Uncommenting the cleanup line does not help, because after a few Fails and 
>>> Errors in the
>>> test suite the script exits immediately. I have not found any way to exit 
>>> the python command
>>> with the test with a nonzero return value that would not cause ‘bash -e’ to 
>>> immediately exit,
>>> so the only way to get rid of the compiled files and still save the test 
>>> result might be to do
>>> the cleanup within the python call itself:
>>> 
>>> %p/bin/python%type_raw[python] -B -c 'import matplotlib as m, os, sys; 
>>> r=m.test(verbosity=1); os.system("find 
>>> %b/build/lib.macosx-*-%type_raw[python] -name \*.py[oc] -exec rm {} \;"); 
>>> sys.exit(1-r)'
>>> 
>>> lets the test phase finish with a warning for the current bunch of 
>>> failures, and should pass if
>>> the tests all succeed. Bit lengthy perhaps, but it sees to do the job.
>>> 
>>> Cheers,
>>>                                     Derek
>>> 
>> 
>> How about putting the command (or perhaps a different one) at the end of the 
>> InstallScript instead and apply it to the deb root directory (%i)?  The 
>> .deb-directory validation is run after the InstallScript.
>> 
>> If the package is building the .deb differently depending on whether the 
>> test suite is run or not, it breaks our fundamental premise of invariant 
>> .debs.
> 
> The cleanup is actually there to make sure it does build the same way, as it 
> only removes the
> files that were byte-compiled in the test run in the first place (since 
> 'python -B' has no effect here
> for whatever reason).
> Putting it into the InstallScript to run on %i at the end would be entirely 
> counter-productive, as it
> would just remove _all_ .pyc files that were precompiled by 'setup.py 
> install' in the right place:
> 
> byte-compiling 
> /scratch.noindex/fink.build/root-matplotlib-py27-1.5.1-2/sw/lib/python2.7/site-packages/mpl_toolkits/tests/test_axes_grid.py
>  to test_axes_grid.pyc
> byte-compiling 
> /scratch.noindex/fink.build/root-matplotlib-py27-1.5.1-2/sw/lib/python2.7/site-packages/mpl_toolkits/tests/test_axes_grid1.py
>  to test_axes_grid1.pyc
> byte-compiling 
> /scratch.noindex/fink.build/root-matplotlib-py27-1.5.1-2/sw/lib/python2.7/site-packages/mpl_toolkits/tests/test_mplot3d.py
>  to test_mplot3d.pyc
> running install_egg_info
> running egg_info
> …
> Writing control file...
> Writing md5sums file...
> Validating .deb dir 
> /scratch.noindex/fink.build/root-matplotlib-py27-1.5.1-2...
> Package looks good!
> 
> One could certainly call the find command _at the beginning_ of Install (and 
> still on %b/build),
> which should do the job as well in perhaps somewhat simpler language. On the 
> other hand I
> find it preferable for the TestScript to cleanup its own mess, and exactly 
> that…
> 
> Cheers,
>                               Derek
> 

Ah, OK, I understand the situation a bit better now.  

Interestingly enough, I didn’t trigger the failure when building 
matplotlib-py35 or matplotlib-py27 here (10.11) using -m mode, possibly because 
of test errors.  I’ll send those in a  separate thread.

-- 
Alexander Hansen, Ph.D.
Fink User Liaison


------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
_______________________________________________
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
List archive:
http://news.gmane.org/gmane.os.apple.fink.devel
Subscription management:
https://lists.sourceforge.net/lists/listinfo/fink-devel

Reply via email to