On 02/21/2014 01:30 AM, Ned Deily wrote:
In article <53070A8A.8080101@hastings.org>,
 Larry Hastings <larry@hastings.org> wrote:
As before you'll find the results here:

    http://midwinter.com/~larry/3.4.status/
Status says that: 

eef7899ea7ab Doc: do not rely on checked-out Sphinx toolchain from 
svn.python.org anymore

is unmerged, which is what Georg and I agreed upon in Issue20661.  Yet, 
in the current python.3.4.2014.02.21.00.07.42.tgz tarball, that change 
appears to be present and, as such, causes installer builds to fail.


Whoopsie!  My local branch is actually correct.  But!  I effectively did this in my automation:
% hg clone 3.4 python-{time}
% cd python-{time}
% rm -rf .hg* .bzr* .git*
% cd ..
% tar cvfz python-{time}.tgz python-{time}
Can you spot the error?  That's right: when you clone, the clone always starts out in "default".  So all the tarballs I've published so far have been wrong.  ARGH.  Sorry!

I've added a "hg branch 3.4" in a judicious spot in my automation.  The resulting tarball doesn't have the sphinx toolchain change.

Look for a new tarball in a few minutes, once "make test" finishes,


/arry