[PATCHv2 2/2] realpath: don't lose errno on failure

[Roland McGrath]

It should be easy enough to write a regression test for this using a
locally defined free.  So I think you don't get any exemption from the
usual policy of regression tests for user-visible bug fixes. :-)

Also, "saved_errno" is the variable name usually used for saving and
restoring errno so as not to touch it.  It would be more clear to name it
something like "failure_errno".


Thanks,
Roland