Hi,

Please see the attached patch which fixes a bug in `format-spec': that a 
format spec like "%>15t", which is intended to limit the width of a 
string to 15 characters, would also pad a string less than that length 
to be that length.

Please note the test case which the patch essentially disables: it calls 
one of the intermediate functions with arguments which I'm not sure it 
would be called with in real usage.  I added two test cases to cover the 
behavior which the patch is intended to fix, and after editing the 
`format-spec--do-flags' to pass the new cases, only that one case 
failed; so, since I'm not sure whether it indicates an actual bug, I 
disabled it.

Perhaps there exists a real-world scenario to which that test would 
apply, but such is not encoded as an end-to-end test of `format-spec' 
with a spec string, so it's hard to say.

In other words, this patch fixes a real bug and adds test cases for it. 
It also causes an existing case to fail, but since I'm not sure whether 
that represents a real bug, that case is now disabled.

All other tests in the file still pass.

Thanks,
Adam