http://docs.python.org/py3k/library/re.html#simulating-scanf

0[xX][\dA-Fa-f]+   ->   (0[xX])?[\dA-Fa-f]+

The documentation appears to be correct to me.  Can you demonstrate your suggestion with some examples?

the prefix "0x" is not necessary for the %x specifier in C
if the pattern will see "ABC", it will not match with it, but it should match

#include <stdio.h>

int main(void)
{
    unsigned n;
    
    scanf("%x", &n);
    printf("%u\n", n);
    return 0;
}


[guest@localhost c]$ .ansi t.c -o t
[guest@localhost c]$ ./t
0xa
10
[guest@localhost c]$ ./t
a
10
[guest@localhost c]$
[guest@localhost c]$ alias .ansi
alias .ansi='gcc -ansi -pedantic -Wall'
[guest@localhost c]$

I checked Standard C by Plauger & Brodie and as I read it, it agrees with py.user and his C compiler. For stdlib strtol() and strtoul(), the 0x/0X prefixes are accepted but optional for explicit base 16. If base is given as 0, they are accepted and set the base to 16 (which is otherwise 10). Except for %i, Xscanf functions apparently call either of the above with an explicit base, which is 16 for the %x specifiers.

the same problem in the %o analog

valid strings for the %x specifier of scanf():
"+0xabc"
"-0xabc"
"+abc"
"-abc"

valid strings for the %o specifier of scanf():
"+0123"
"-0123"
"+123"
"-123"

how to patch
the %x specifier:
0[xX][\dA-Fa-f]+   ->   [-+]?(0[xX])?[\dA-Fa-f]+
the %o specifier:
0[0-7]*            ->   [-+]?[0-7]+

New changeset b26471a2a115 by Ezio Melotti in branch '2.7':
#14519: fix the regex used in the scanf example.
http://hg.python.org/cpython/rev/b26471a2a115

New changeset e317d651ccf8 by Ezio Melotti in branch '3.2':
#14519: fix the regex used in the scanf example.
http://hg.python.org/cpython/rev/e317d651ccf8

New changeset 7cc1cddb378d by Ezio Melotti in branch 'default':
#14519: merge with 3.2.
http://hg.python.org/cpython/rev/7cc1cddb378d

Fixed, thanks for the report and the suggestions!