This is the mail archive of the cygwin-developers 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]
Other format: [Raw text]

Re: More: [1.7] packaging problem? Both /usr/bin/ and /usr/lib/ are non-empty


On Wed, May 13, 2009 at 10:56:58AM +0200, Corinna Vinschen wrote:
>On May 12 23:02, Christopher Faylor wrote:
>> On Tue, May 12, 2009 at 11:41:16AM +0200, Corinna Vinschen wrote:
>> >Ok, here's another patch.  [...]
>> 
>> Wouldn't it be better to check entries as fstab is being parsed rather
>> than rescan the whole table a second time?  It seems like just adding a
>> couple of strcmp's as in add_itm would be all that you need.
>
>Fine with me.
>
>> Also, since mount already has a force entry maybe this option should
>> be called "override" instead.  And, I agree with Eric that calling this
>> "read-only" is confusing.  It doesn't mean what you'd expect if you have
>> an idea of a similar option from linux.
>
>Yes, I had already changed that locally.  Really, I have no strong
>opinion for the exact words.  I would just like to have this sorted out
>the way we discussed it.
>
>> With those ideas in mind, I present my variation of your patch.
>> 
>> I've called what we do with '/' "immutable".  An immutable mount setting
>> needs to be overridden with an "override" option.
>> 
>> I've also marked all of the automatic mounts with a ",auto" to make it
>> clear that Cygwin is creating them automatically.  I also added this
>> to cygdrive mounts.
>
>fillout_mntent shouldn't do that.  "auto" and "immutable" are internal
>flags, not valid mount options.  If fillout_mntent adds them to mnt_opts
>they will be printed by mount(1), which is not a good idea if you use
>the `mount -m' command to generate an /etc/fstab output.

I thought it was nice to be able to see which files came into being
because of a Cygwin decision and which were called for specifically.  Linux
adds "stuff" in that field which wasn't specified on the command line.

But, regarding, "mount -m": It looks like more mount work is required
there since we don't want mount -m to generate mount commands that will
fail (as in the case of root) or to force a mount of /usr/lib when it
isn't necessary.  The ",auto" would be a nice clue about that.  We could
just have mount ignore that option like linux's mount does with some
options that show up in its mount output.

I thought it might be nice to make the "immutable" option a real option
at some point, recognized by mount and fstab.

>> Oh, and one additional thing that I did was allow the use of -o nouser
>> as a mount option.  Was there a reason why we disabled that for 1.7?
>
>Yes.  The nouser option doesn't make sense for the mount command because
>you can't override a nouser mount point with the mount command anyway.
>Every mount point created by mount is a user mount point.

That is exactly what I changed.  Since linux doesn't have the concept of
a "user" mount, it seems counterintuitive not to be able for a
sufficiently privileged user to be able to update a "system" mount
without needing to change /etc/fstab and restarting every Cygwin process.

cgf


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