Tim Hatch wrote: > They were introduced in r9037 with the unconditional call to realpath, > and most of them were easy enough to work around (done in r9046). One, > however, I left alone[2] because it doesn't smell right. This one > appears to be calling realpath too early, and ending up with the wrong > directory. Remy, can you please check?
In that last case, the test case was wrong, as it assumed that os.path.dirname(base) returns the parent directory (even for symlinks). I have fixed the test cases slightly differently by taking the realpath() of the temporary directory in setUp() instead of calling realpath() in all test cases: http://trac.edgewall.org/changeset/9047 -- Remy
signature.asc
Description: OpenPGP digital signature
