Hi Carlos, As part of my work on Ubuntu, I have updated the feedparser package to include a python3-feedparser binary package. Along the way, I also upgraded to feedparser 5.1, and carried over our Ubuntu delta to use dh_python2. I also updated the package to compat level 8, and simplified the rules file. I understand that these changes may be too radical, but I do think it ultimately makes supporting Python 3 easier. I see that you have also updated to 5.1 in DPMT svn, so some of my work duplicates that, but not too much. As you may know, feedparser 5.1 supports Python 3, but let's just say there are a few problems ;) which I've solved in my package. * The test suite has several failures, some of which only show up when chardet is installed (as it is a build-dep). When building upstream 5.1 in a virtualenv without chardet, these don't fail. I disable these tests in a quilt patch. This bug has been reported upstream. * Some of the parse_date_rfc822 tests do not overflow as expected by the test suite on some platforms. I disable these tests. This bug has been reported upstream. * test_unicode_2 fails in an sbuild because sys.getfilesystemencoding() is us-ascii for me at least. When running the tests in a normal environment on upstream 5.1, this passes because the file system encoding is utf-8. Anyway, I also disable this test. This bug has been reported upstream. * There are lots of problems running feedparsertest.py in Python 3, so I just don't :). You would have to copy it to some unique place, run it through 2to3, then arrange to have feedparser/test be findable by it. It didn't seem worth it, but I did test the resulting binary package in a chroot and it seems to work. * The upstream comes with sgmllib3.py which is a 2to3'd version of Python 2's sgmllib.py. This makes up for the removal of sgmllib from Python 3. I didn't like the solution recommended by upstream to handle this, so I install the file as feedparser_sgmllib3.py and quilt patch feedparser.py to import that as sgmllib (with proper handling if it can't be found). * The upstream 5.1 tarball is missing some data files used during the test suite. I add those back and use a debian/source/include-binaries file to make things happy. This bug has been reported upstream. * Funny enough, the python3-feedparser package triggers the embedded-feedparser-library lintian warning, which I override. Once this lands, I'll open a bug on lintian to avoid that warning, since it's clearly bogus in this case. I think that's about it. I'd like to ask you what is the best way to get you the changes. I can open a bug and attach a patch to that, or I can stick something in DPMT svn (e.g. .../feedparser/branches/py3 or somesuch). What is your preference? Cheers, -Barry
Attachment:
signature.asc
Description: PGP signature