_winreg.EnumValue causes MemoryError

[kzmi]

BPO	9937
Nosy	@vstinner, @tjguk, @eryksun, @iritkatriel
Files	fonts.reg: Sample registory data test.py: py file to reproduce

^{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 = None
closed_at = <Date 2022-01-07.10:06:55.079>
created_at = <Date 2010-09-24.09:17:15.592>
labels = ['type-bug', 'OS-windows']
title = '_winreg.EnumValue causes MemoryError'
updated_at = <Date 2022-01-07.10:06:55.076>
user = 'https://bugs.python.org/kzmi'

bugs.python.org fields:

activity = <Date 2022-01-07.10:06:55.076>
actor = 'eryksun'
assignee = 'none'
closed = True
closed_date = <Date 2022-01-07.10:06:55.079>
closer = 'eryksun'
components = ['Windows']
creation = <Date 2010-09-24.09:17:15.592>
creator = 'kzmi'
dependencies = []
files = ['18993', '18994']
hgrepos = []
issue_num = 9937
keywords = ['patch']
message_count = 14.0
messages = ['117264', '117265', '117266', '117289', '117291', '117308', '117310', '117403', '117404', '117438', '117439', '222618', '409943', '409956']
nosy_count = 6.0
nosy_names = ['vstinner', 'ocean-city', 'tim.golden', 'kzmi', 'eryksun', 'iritkatriel']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue9937'
versions = ['Python 2.7', 'Python 3.4', 'Python 3.5']

[kzmi]

_winreg.EnumValue causes MemoryError if the name of the value contains multibyte unicode characters.

I'm using 2.6.6.
I think it is relevant to Isssue bpo-2810.

[kzmi]

Added sample registory data file.

[kzmi]

Added py file to reproduce.

[briancurtin]

I'm not able to reproduce any MemoryError on 2.6, 2.7, or 3.2 on Windows 7 using your registry data and test script.

[ocean-city]

I can reproduce the issue on python2.x. I cannot on
python3.x.

[ocean-city]

This happens because RegQueryInfoKeyA returns maximum value
name length not in *Ansi* characters but in *Unicode" characters.
I don't know clue why.

http://msdn.microsoft.com/en-us/library/ms724902%28VS.85%29.aspx

In multibyte environment, 2 or 3 bytes characters are
mapped to one size unicode. So usually, the length of *Ansi*
characters are larger than or equals to the length of *Unicode*
characters. If every Ansi characters are mapped to one size
Unicode, lengthes are equal, so probably this issue won't happen.

I'll attach the patch to fix this issue. But be care, this is not
tested so densely. At least, the script to reproduce this issue
ran fine.

[ocean-city]

Again, I'm not sure this is right fix.

[ocean-city]

Sorry, my patch didn't pass test_dynamic_key(test_winreg).
(Apologize for not testing on unit test)

RegEnumValueA and RegQueryInfoKeyA always returned 0 for
required data size if target is HKEY_PERFORMANCE_DATA.

The possible fixes I can imagine are...

1. Use GetCPInfo() to get MaxCharSize.
   http://msdn.microsoft.com/en-us/library/dd317780%28v=VS.85%29.aspx
   We can get a large value enough to store the key name in ANSI by
   "The length in Unicode" * MaxCharSize.

2. Use RegEnumValueW and convert result via WideCharToMultiByte

I'll attach the patch with option 1.

[ocean-city]

I shall create unit test later.

[ocean-city]

I've withdrew my patch... It has still problem, and
I cannot fix it yet. I suppose option 2 is easy.
ANSI API's limitation makes it harder to implement
correctly.

[vstinner]

Can't we use RegEnumValueW and RegQueryInfoKeyW?

[BreamoreBoy]

Does anyone wish to follow this up as I no longer run 2.7?

[iritkatriel]

I can reproduce the issue on python2.x. I cannot on python3.x.

Did anyone reproduce on python3.X? Why are versions 3.4/3.5 selected?

[eryksun]

Did anyone reproduce on python3.X?

In principle, winreg should not have a problem in 3.x because it only uses the UTF-16 wide-character API.