Update os.pread to accept the length type as size_t

[corona10]

BPO	38265
Nosy	@vstinner, @serhiy-storchaka, @corona10
PRs	bpo-38265: Update os.pread to accept the length type as Py_ssize_t #16359

^{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 2019-09-25.11:14:19.644>
created_at = <Date 2019-09-24.18:14:15.456>
labels = ['library', '3.9']
title = 'Update os.pread to accept the length type as size_t'
updated_at = <Date 2019-09-25.11:53:20.902>
user = 'https://github.com/corona10'

bugs.python.org fields:

activity = <Date 2019-09-25.11:53:20.902>
actor = 'serhiy.storchaka'
assignee = 'none'
closed = True
closed_date = <Date 2019-09-25.11:14:19.644>
closer = 'vstinner'
components = ['Library (Lib)']
creation = <Date 2019-09-24.18:14:15.456>
creator = 'corona10'
dependencies = []
files = []
hgrepos = []
issue_num = 38265
keywords = ['patch']
message_count = 11.0
messages = ['353107', '353108', '353109', '353110', '353111', '353143', '353144', '353166', '353172', '353174', '353183']
nosy_count = 3.0
nosy_names = ['vstinner', 'serhiy.storchaka', 'corona10']
pr_nums = ['16359']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue38265'
versions = ['Python 3.9']

[corona10]

I've found this comments.
// TODO length should be size_t! but Python doesn't support parsing size_t yet.
(

cpython/Modules/posixmodule.c

Line 8830 in 279f446

# TODO length should be size_t! but Python doesn't support parsing size_t yet.

)

I don't know when this comment was created.
But at this point, it seems to provide that functionality.
It can be simply solved through the clinic.

How about updating this argument to be size_t and remove the TODO comment?

[serhiy-storchaka]

Length should be Py_ssize_t. Python does not support creating bytes objects larger than PY_SSIZE_T_MAX.

[corona10]

Oh okay!
Thanks for letting me know.
Is it okay to switch it into Py_ssize_t?

[serhiy-storchaka]

I think it is okay. Just check that all platforms which support pread() (Linux, *BSD, macOS) pass the length as size_t instead of int. Windows does not always support length larger than 2 KiB, but it does not support pread() either.

[corona10]

Linux: http://man7.org/linux/man-pages/man2/pread.2.html
BSD: https://www.freebsd.org/cgi/man.cgi?pread(2)
macOS: https://developer.apple.com/library/archive/documentation/System/Conceptual/ManPages_iPhoneOS/man2/pread.2.html

All of them pass the length as size_t instead of int.
I will send the PR soon :)

[serhiy-storchaka]

New changeset ad7736f by Serhiy Storchaka (Dong-hee Na) in branch 'master':
bpo-38265: Update os.pread to accept the length type as Py_ssize_t. (GH-16359)
ad7736f

[corona10]

Thank you @serhiy.storchaka and Kyle Stanley for the code review and approve.

This issue is now solved.
I close this issue.

[vstinner]

Python 3.7 and 3.8 also have os.pread(). Why not also fixing these branches?

--

os.pwrite() uses Py_buffer for its argument and supports size_t size.

os.preadv() and os.pwritev() use internally the C structure "struct iovec" which uses size_t for its iov_len field.

[serhiy-storchaka]

Because this looks rather as a new feature than a bug fix. We don't add new features in maintained versions.

Always there is a risk of introducing a regression. We don't know what bugs are contained in 64-bit size support on different platforms. I would not be surprised if for example some old macOS versions accept size_t, but work correctly only with the 32-bit size.

[vstinner]

Because this looks rather as a new feature than a bug fix. We don't add new features in maintained versions.

Oh ok. That's fair. I close the issue.

The workaround is to use os.preadv() or other functions which accept size_t.

[serhiy-storchaka]

Or call pread() multiple times and concatenate results.