struct.pack() on 64bit architectures

[pelotas]

BPO	925932
Nosy	@tim-one, @loewis
Files	patch-Lib_test_test_fcntl_py

^{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 2007-04-10.16:45:14.000>
created_at = <Date 2004-03-30.10:55:25.000>
labels = ['tests']
title = 'struct.pack() on 64bit architectures'
updated_at = <Date 2007-04-10.16:45:14.000>
user = 'https://bugs.python.org/pelotas'

bugs.python.org fields:

activity = <Date 2007-04-10.16:45:14.000>
actor = 'zseil'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['Tests']
creation = <Date 2004-03-30.10:55:25.000>
creator = 'pelotas'
dependencies = []
files = ['5900']
hgrepos = []
issue_num = 925932
keywords = ['patch']
message_count = 4.0
messages = ['45688', '45689', '45690', '45691']
nosy_count = 4.0
nosy_names = ['tim.peters', 'loewis', 'zseil', 'pelotas']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue925932'
versions = []

[pelotas]

I'm running python on OpenBSD/sparc64 (SUN Ultra 10).

On this machine struct.pack() gives me:

>>> struct.pack('l', 1)
'\x00\x00\x00\x00\x00\x00\x00\x01'
>>> struct.pack('i', 1)
'\x00\x00\x00\x01'

On i386 box I have:

>>> struct.pack('l', 1)
'\x01\x00\x00\x00'
>>> struct.pack('i', 1)
'\x01\x00\x00\x00'

Because of this, OpenBSD port uses attached patch. I
guess that this is generic problem (not OpenBSD
specific) and every 64bit platform would suffer.

Am I right?

[tim-one]

Logged In: YES
user_id=31435

For a Unix-head reviewer: the patch is to test_fcntl.py,
which uses a native struct.pack() to build a lockdata
argument for fcntl. On the OP's box, it constructs stuff of
the wrong size, due to "l" format codes producing 8-byte
thingies. The patch changes the "l" codes to "i" on the OP's
platform.

I wonder whether we couldn't instead use "i" codes on all BSD-
ish platforms -- on 32-bit boxes, "i" is synonymous with "l",
while on 64-bit boxes it looks like "l" is wrong but "i" is correct.

[loewis]

Logged In: YES
user_id=21627

I don't quite understand the current code: struct flock is
on Darwin

struct flock {
        off_t   l_start;        /* starting offset */
        off_t   l_len;          /* len = 0 means until end
of file */
        pid_t   l_pid;          /* lock owner */
        short   l_type;         /* lock type: read/write,
etc. */
        short   l_whence;       /* type of l_start */
};

and off_t is

typedef int64_t quad_t;
typedef quad_t off_t; /* file offset */

so it appears there is no padding at all. In any case,
this needs to be changed in posixfile._posixfile_.flock
accordingly.

[zseil]

test_fcntl was fixed in r38766 and r38767 to support
64bit BSD systems:

http://svn.python.org/view?rev=38766&view=rev

That fix is included in Python 2.4.2 and above.
Thanks for the patch anyway!