Python 2.7.7 regression in mimetypes module on Windows

[foom]

BPO	21652
Nosy	@loewis, @jaraco, @vstinner, @tjguk, @benjaminp, @ned-deily, @merwok, @bobince, @bitdancer, @briancurtin, @shimizukawa, @Fak3
Files	finduni.py 21652.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/tjguk'
closed_at = <Date 2014-06-29.20:04:02.457>
created_at = <Date 2014-06-03.14:42:12.003>
labels = ['type-bug', 'library', 'release-blocker']
title = 'Python 2.7.7 regression in mimetypes module on Windows'
updated_at = <Date 2014-06-29.20:04:39.226>
user = 'https://bugs.python.org/foom'

bugs.python.org fields:

activity = <Date 2014-06-29.20:04:39.226>
actor = 'benjamin.peterson'
assignee = 'tim.golden'
closed = True
closed_date = <Date 2014-06-29.20:04:02.457>
closer = 'python-dev'
components = ['Library (Lib)']
creation = <Date 2014-06-03.14:42:12.003>
creator = 'foom'
dependencies = []
files = ['35620', '35795']
hgrepos = []
issue_num = 21652
keywords = ['patch']
message_count = 14.0
messages = ['219693', '220353', '220383', '220455', '220458', '220462', '220466', '220469', '220496', '220539', '221314', '221783', '221890', '221892']
nosy_count = 29.0
nosy_names = ['loewis', 'exarkun', 'jaraco', 'foom', 'vstinner', 'tim.golden', 'benjamin.peterson', 'ned.deily', 'eric.araujo', 'kaizhu', 'aclover', 'r.david.murray', 'brian.curtin', 'Suzumizaki', 'frankoid', 'Dmitry.Jemerov', 'shimizukawa', 'vldmit', 'Vladimir.Iofik', 'Vladimir Iofik', 'python-dev', 'Roman.Evstifeev', 'adamhj', 'me21', 'Hugo.Lol', 'Daniel.Szoska', 'Micha\xc5\x82.Pasternak', 'quick.es', 'gcarothers']
pr_nums = []
priority = 'release blocker'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue21652'
versions = ['Python 2.7']

[foom]

The change done for bpo-9291 in Python 2.7.7 caused the mimetypes.types_map dict to change to contain a mixture of str and unicode objects, rather than just str, as it always had before.

This causes twisted.web to crash when serving static files on Windows. See https://twistedmatrix.com/trac/ticket/7461 for the bug report against Twisted.

[gcarothers]

Issue also exists in Pyramid (any wsgi server/framework) See Pylons/pyramid#1360 for Pyramid bug.

[ned-deily]

If there is a regression to be fixed, there needs to be a patch with a test. Anyone?

[foom]

Reverting the incorrect commit which caused the regression would be a good start.

[bitdancer]

If I understand correct that patch was itself attempting to fix a regression ;)

[tjguk]

Only if you have something better to put in its place! That commit was
fixing an existing problem; perhaps not your problem, but someone's. To
revert it would simply move the pain around.

I hope to be able to work on this fairly soon. If anyone else wants to
propose an approach to take this forward, they're welcome to do so.

[exarkun]

That commit was fixing an existing problem; perhaps not your problem, but someone's. To revert it would simply move the pain around.

Doesn't the very same logic apply to the original commit?

I hope to be able to work on this fairly soon.

Thanks. Much appreciated.

[VladimirIofik]

Did anyone here managed to reproduce the issue?
I tired but I have failed to.
I'm still trying though.

Anyway I'd start from investigating why a function EnumKey returns Unicode object while according to documentation it should return string.

[loewis]

I can reproduce the issue with the attached script. The problem is not that EnumKey returns a Unicode string (which it never does), but that QueryValueEx returns a Unicode string. See http://hg.python.org/cpython/rev/18cfc2a42772 for the actual change in question.

[VladimirIofik]

Thanks, Martin. I didn't read the final with enough care. I think the second part of changes (the ones that are around QueryValueEx) should be rolled back. I believe the code was correct at that part.
Tim, what do you think?

[benjaminp]

Vladimir, if you could provide a patch implementing your proposed rollback, that would be appreciated.