Hello,
at the moment, fnmatch.fnmatch() will fail to match any string, which
has \n character. This of course breaks glob as well.

Example

> import fnmatch
> fnmatch.fnmatch("foo\nbar", "foo*")
False

> import glob
> open("foobar", "w").close()
> open("foo\nbar", "w").close()
> glob.glob("foo*")
['foobar']

while the expected result is ['foobar', 'foo\nbar']. The standard C
fnmatch function from fnmatch.h is behaving correctly i.e. this code
will print out "match!"

#include <fnmatch.h>
#include <stdio.h>

int main()
{
    if (fnmatch("foo*", "foo\nbar", FNM_NOESCAPE) == 0)
        printf("match!\n");
    else
        printf("fail!\n");
    return 0;
}


This misbehaviour is caused by the fnmatch.translate() which adds $ to
the end of the regexp. Without the ending $ the fnmatch function works OK.

haha wow I just read the fnmatch code... trunk r2734 | guido | 1992-01-12 

added fnmatch._cache for it to cache compiled regular expressions.  That
has -long- since become unnecessary as the re module does that itself. ;)

I'll clean this up while fixing this bug.

aww, i guess the _cache does cache the result of the glob -> regular
expression translation.  it needs to stay for that.

changing the '$' to \Z(?ms)' fixes the problem.

fixed in trunk r74475, py3k r74476

I'll backport this to 2.6 and 3.1 later (its too late for 3.1.1).

release26-maint r76023 to appear in Python 2.6.5.
release31-maint r76024 to appear in Python 3.1.2.