Re: cygwin 1.7: why there is a directory E/cygwin/dev in the tree of cygwin ?

[Eric Blake <eblake at redhat dot com>]

[reviving an old thread, relevant to today's current bash postinstall
failures]

On 03/17/2010 12:28 PM, Corinna Vinschen wrote:
> On Mar 17 12:21, Eric Blake wrote:
>> On 03/17/2010 02:19 AM, rolandc wrote:
>>> I do not understand why the postinstall script bash.sh is so complex
>>>
>>>     DEVDIR="$(cygpath -au "C:/$(cygpath -am /dev/)" | sed 's|/c/\(.\):/|/\1/|')"
>>>     mkdir -p "$DEVDIR" || result=1
>>>
>>> it would be simple (too simple?) to
>>>     mkdir -p /dev || result=1
>>
>> Yes, it would be too simple.  /dev already exists, so the mkdir would
>> fail to do anything useful.  We REALLY want to create the underlying
>> Windows directory at the same location at where /dev would be mounted,
>> and to do that, we really do want to know the windows location (drive
>> letter and all) of /.  Then, by using mkdir of that fancy windows path
>> that happens to live at the same place as where /dev normally resolves
>> to, then we can guarantee that /dev/stdin gets created as an actual
>> symlink in the windows heirarchy (since it does NOT resolve via the /dev
>> magic mount point), and that tab-completion can see any contents placed
>> into the windows counterpart directory.
> 
> Nothing of this should be necessary since the 000-cygwin-post-install.sh
> script from the base-cygwin package already creates /dev.

Interesting - cygwin 1.7 is much nicer in regards to letting 'mkdir
/dev' succeed without having to go through cygpath hoops.

I'm building a new bash package now that should fix all this mess, by
using the same means as 000-cygwin-post-install.sh to populate necessary
entries into /dev.

-- 
Eric Blake   eblake@redhat.com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature