Python-2.5

[dnicholson@…]

Now version 2.5.

​http://www.python.org/download/releases/2.5/

[bdubbs@…]

Current version is now 2.5. It also has the security fix.

​http://www.python.org/download/ has info

[Randy McMurchy]

Be it known that I am against using the 2.5 version in BLFS-6.2.0. There are API changes that affect code and could affect Python's interaction with other packages. Please review ​http://docs.python.org/whatsnew/porting.html

My understanding is 2.4.4 fixes all the security issues without introducting API changes.

My observations are based on the fact that Python is such a heavy dependency on many, many packages.

I could be wrong, but is there really time to test the API changes in the 2.5 series?

[bdubbs@…]

I can update to 2.4.4 or 2.5. We are currently using Python 2.4.3 on quantum for Trac.

Your call.

[Randy McMurchy]

I will not put myself in the position of "making the call". It will be a joint/community decision or a single editor's decision. If that makes me an incompentent person to make decisions, then so be it.

Bruce, you are that single Editor. So, it is actually your call.

I will say this:

If it were me, I'd go with 2.4.4 as long as the information I'm interpreting at ​http://www.python.org/news/security/ is valid information. According to this, the 2.4.4 version fixes all known security problems. Again, I could be wrong in my interpretation.

I am concerned about the API changes. Python is practically a mandatory dependency for many packages, including the GNOME desktop. The 2.4.x series is *known* to work. 2.5 will be a mystery. There simply isn't time to test everything that depends on Python.

Just my thoughts, anyway. Hopefully, others will chime in. Bruce, if you're confident with 2.5, then do it. I can't offer much more than that.

[Randy McMurchy]

I suppose I should add that my observations are based on the fact that there has been very few BLFS updates in the last several months. To throw an API change with a major package into the picture seems risky.

If things were as they were a while ago, where the BLFS packages are semi-current, I would think differently. But right now, I feel that BLFS has been stagnant for too long to throw API changes into the picture.

[bdubbs@…]

I understand your concerns. Since I don't use Gnome, I really can't test that. 2.4.4 it is.

[dnicholson@…]

Just to add my .02, I agree with Randy. I've got Python-2.4.4 over here with a whole bunch of stuff stacked upon it in gnome and haven't seen any issues yet.

[bdubbs@…]

I've commited Python 2.4.4 but will keep this ticket open for Python 2.5 and retarget it for future.

[matthew@…]

I've got issues with Python-2.5's testsuite.

Summary:

276 tests OK.
4 tests failed:
    test_shutil test_subprocess test_tarfile test_tempfile

Details:

test_shutil
test test_shutil failed -- errors occurred in test.test_shutil.TestShutil

test test_subprocess failed -- errors occurred in test.test_subprocess.ProcessTestCase

test_tarfile
test test_tarfile crashed -- <type 'exceptions.OSError'>: [Errno 2] No such file or directory: '/tmp/testtar.dir'

test_tempfile
Exception exceptions.AttributeError: "mkstemped instance has no attribute 'fd'" in <bound method mkstemped.__del__ of <test.test_tempfile.mkstemped instance at 0x95e182c>> ignored
Exception exceptions.AttributeError: "mkstemped instance has no attribute 'fd'" in <bound method mkstemped.__del__ of <test.test_tempfile.mkstemped instance at 0x95e12ac>> ignored
Exception exceptions.AttributeError: "mkstemped instance has no attribute 'fd'" in <bound method mkstemped.__del__ of <test.test_tempfile.mkstemped instance at 0x95e12ac>> ignored
Exception exceptions.AttributeError: "mkstemped instance has no attribute 'fd'" in <bound method mkstemped.__del__ of <test.test_tempfile.mkstemped instance at 0x95e12ac>> ignored
test test_tempfile failed -- errors occurred; run in verbose mode for details

I think the tar and tempfile test failures are actually linked. The really worrying thing is that my entire /tmp directory is deleted during those tests!

I'd really appreciate it if someone could confirm that this also happens to them (though obviously you might want to make sure you have nothing valuable in /tmp first!).

[Randy McMurchy]

Cannot confirm. The tests ran perfectly. Using LFS-SVN-20070210. Here's the results:

test_syntax test_sys test_tarfile test_tcl test_tempfile test_textwrap test_thread . . . test_zipimport test_zlib 282 tests OK. 37 tests skipped:

test_aepack test_al test_applesingle test_bsddb test_bsddb185 test_bsddb3 test_cd test_cl test_codecmaps_cn test_codecmaps_hk test_codecmaps_jp test_codecmaps_kr test_codecmaps_tw test_curses test_gl test_imgfile test_linuxaudiodev test_macfs test_macostools test_nis test_normalization test_ossaudiodev test_pep277 test_plistlib test_scriptpackages test_socket_ssl test_socketserver test_sqlite test_startfile test_sunaudiodev test_timeout test_unicode_file test_urllib2net test_urllibnet test_winreg test_winsound test_zipfile64

1 skip unexpected on linux2:

test_bsddb

[Randy McMurchy]

Unassigning myself from the ticket as I'm not qualified to examine nor determine the need for the patch Dan spoke of. I'll let someone else do this.

See ​http://linuxfromscratch.org/pipermail/blfs-dev/2007-February/016690.html

[jciccone@…]

In that link, ​http://www.paldo.org/paldo/sources/Python/Python-2.5-build-fix-1.patch.bz2 is a multilib hack and is not needed for single arch builds. I prefer to just have 2 Python installations and switch between them with an enviornment variable.

[dnicholson@…]

I don't think the one on paldo is a multilib hack since paldo doesn't support multilib. I'm gonna email Jürg and see what he says.

[dnicholson@…]

Adding a couple data points after doing a build today. As mentioned somewhere on blfs-dev, the paldo patch was needed to get the --enable-shared build working when there wasn't an existing Python installed. No one here has had that issue, and I didn't have that issue today on a fresh build. I'd say we can move along without it.

Also, all the tests passed for me. Actually, two of the network ones bombed at first because I hadn't properly initialized /etc/hosts yet.

[dnicholson@…]

Crap. Scratch that. That was all using Python-2.4.4. I don't know if the same applies for 2.5.

[Randy McMurchy]

Updated BLFS to Python-2.5

[(none)]

Milestone 6.2.1 deleted