This is the mail archive of the cygwin 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: mkdir -p and network drives

On Thu, May 05, 2005 at 06:44:05PM +0100, Dave Korn wrote:
>----Original Message----
>>From: Igor Pechtchanski
>>Sent: 05 May 2005 18:20
>> On Thu, 5 May 2005, Paul Eggert wrote:
>>> Eric Blake <ebb9@XXX.XXX> writes:
>>>> //MACHINE currently generates ENOENT, whether or not there is a
>>>> server on the network with that name, and mkdir(2), stat(2), and
>>>> chdir(2) with an argument of "//MACHINE" fail.
>>> That's certainly a hassle.  Let's not worry about going through
>>> zillions of lines of application code working around that
>>> incompatibility.  Let's get Cygwin fixed instead.
>> One potential problem is that on Windows, using NetShareEnum when checking
>> the path will mean that every time someone does "mkdir -p", the command
>> will have to wait until all network shares are examined.  This could
>> result in quite a bit of a delay, to say the least...
>Um, I don't see why.  You only need to check network shares at all when
>you have "//" at the start of the command line.  I can't see any reason
>why you'd want to check shares on any machine other than //MACHINE
>either, and in fact I can't see why you'd care about checking any share
>on //MACHINE other than the one specified after '//MACHINE/' in the
>command line.  Could you perhaps expond on that statement a bit?

I don't know precisely how mkdir works but if it does something like:

1) Does '//' exist?
2) Does '//MACHINE' exist?
3) Does '//MACHINE/x' exist?

Step 1 would work since '//' is the root directory.  Step 2 could be
looked up as a hostname.  Step 3 could be determined via normal filename
tests.  I don't see why it should be necessary to iterate over every
available share either.

The problem is that //MACHINE isn't really a directory.  You can't cd to
it.  You can't create stuff under it.  If it was really going to be supported,
we could make cygwin treat this as a special case but, as I said, this isn't

To *really* fully support '//' would be interesting.  Should all of the
machines in the network show up if you do a "ls //"?  I think not,
although it would be sort of nifty to be able to cd to some directory
and get a listing of all of the available machines on your network.


Unsubscribe info:
Problem reports:

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