This is the mail archive of the cygwin@sourceware.cygnus.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

File permissions problems


List:
  Eric Fifer & I are attempting to resolve the remaining problems in
perl5.5.6x0 (under cygwin) in preparation for the official release of a
stable perl5.6. I'm not sure what the differences in his setup and my
setup are, but I'm getting more errors that he. A few of them are weird,
and seem directly related to underlying problems in the cygwin dll,
especially ntsec. For example, I get failures in the following tests:

Failed Test  Status Wstat Total Fail  Failed  List of failed
-------------------------------------------------------------------------------
base/rs.t        13  3328    14   14 100.00%  1-14
io/pipe.t         1   256    15   12  80.00%  7-15
lib/anydbm.t                 12    4  33.33%  1-2, 4, 9
lib/dirhand.t                 5    1  20.00%  3
lib/filecache.t 255 65280     1    1 100.00%  1
lib/filepath.t  255 65280     4    4 100.00%  1-4
lib/glob-basic.               9    1  11.11%  2
lib/io_dir.t                 10    1  10.00%  3
lib/io_pipe.t                10    2  20.00%  9-10
lib/io_poll.t                 8    1  12.50%  4
lib/io_unix.t     0    15     5    4  80.00%  2-5
lib/ipc_sysv.t               16   16 100.00%  1-16
lib/ndbm.t       13  3328    64   50  78.12%  1-2, 4, 9,
19-64
lib/odbm.t        9  2304    64   64 100.00%  1-64
lib/posix.t                  26    3  11.54%  8-10
lib/sdbm.t       13  3328    66   50  75.76%  1-2, 4, 9,
21-66
op/grent.t                    1    1 100.00%  1
op/magic.t                   35    2   5.71%  3, 5
op/stat.t                    58    9  15.52%  2, 18-21,
26-27, 57-58
op/taint.t                  149    3   2.01%  1, 3, 31
pragma/warnings             333    1   0.30%  117
4 tests and 29 subtests skipped.
Failed 21/234 test scripts, 91.03% okay. 228/11364 subtests
failed, 97.99% okay.

I'm just going to highlight a few of these errors. The base/rs.t failure
is specifically reported by perl harness as a "permission denied" error.
Eric suggested that some problems may be related to the permissions of
my /tmp, since the lib/io_unix.t and lib/ipc_sysv.t (via ipc-daemon)
access files there.  Here's the result of my investigation:

/tmp is drwxrwxrwx  cwilson Users
umask is 002

However, inside /tmp there are a number of files whose permissions seem
messed up:

MultiFileMsg
MultiFileSem
MultiFileShm
sock-11796
sock-13111

I can't even get a 'ls -l' on those files, without a permission denied
error. The windows Properties->Security->Permissions dialog on the
Multi* files shows:

"The security information for F:\cygnus\cygwin-pre21\tmp\MultiFileMsg is
not standard and cannot be displayed. Windows NT 3.x and 4.x support
certain features such as Deny Access Control Entries but cannot edit
ssecurity information which uses these features. The information may
have been modified by a computer running Windows NT 5.0, which supports
these features and can edit information which uses the. Do you want to
overwrite the current security information?"

The same dialog on the sock-* files shows:

Administrators -- Special Access (O)
cwilson -- Special Access (PO)
Everyone -- Special Access
Users -- Special Access

The Multi* files are created by ipc-daemon, and I guess the sock-* files
are created by the io_sock test (via the AF_UNIX cygwin subsystem.) In
any case, this looks like a cygwin problem. BTW, Windows2000/NT5 has
never been anywhere near my computer.

I thought that the cygwin layer was supposed to make all the permissions
stuff "transparent" -- e.g. just like unix -- so that individual
applications like cygipc didn't need to do anything special in order to
get the expected permissions behavior; and so that individual
applications couldn't foobar the permissions by trying to fiddle with
them directly. So, it seems that if the permissions 'bits' got scrogged,
then cygwin1.dll must have done it.

FWIW, I'm using the cygwin-inst from Jan 31, and the cygwin1.dll from
Feb 4.
uname -a reports:
CYGWIN_NT-4.0 BELGARION 1.1.0(0.16/3/2) 2000-02-05 01:03:15 i586 unknown

CYGWIN=binmode tty ntea ntsec mixed 
(yeah, I know 'mixed' has no effect anymore; it's leftover from when I
was using Sergey's coolview)

Any ideas as to what could be going wrong here? 

--Chuck

--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]