On 14/12/06, Ilya Zakharevich <[EMAIL PROTECTED]> wrote:
The massive majority of test failures of 5.8.8 on OS/2 turn out to be
false positives. The following patch reduces the number of failing
subtests to 2; both (magic.t and syslog.t) seem to indicate some real
problems with the port.
[I suspect that the goof in t/Command.t was so obvious, that it may
be fixed already.]
Thanks, applied to bleadperl as change #29578, except the change to
getppid.t, since the test file has changed too much. However I'll see
if I can make the same kind of change to the new test.
./lib/ExtUtils/t/Command.t was not binmode()ing a test file
./os2/OS2/ExtAttr/t/os2_ea.t JFS not allowing EAs with certain chars;
we do not test for bugs in
filesystems, so avoid these chars
./t/io/fs.t fchown() may call chown(), so the error
message may come from chown()
./t/op/getppid.t Reparenting semantic different on OS/2:
ppid() goes up the tree, not to 1
./ext/IO/t/io_unix.t On OS/2, "unix domain"-like sockets live
in a separate part of the
filesystem; we relocate those
with absolute path there,
but relative-path ones are
not supported
./t/op/stat.t Was living garbage behind
./lib/Time/Local.t Was assuming that time_t is signed
