lyx has problem with network directory names

[Larry Hall (Cygwin)]

Enrico Forestieri wrote:
> Larry Hall writes:
> 
>> Enrico Forestieri wrote:
>>> David Arnstein writes:
>>>> I just tried to use the Cygwin port of lyx. It cannot cope with my
>>>> home directory, which appears as //fs-sj1-15/darnstein in Cygwin. This
>>>> is a network directory, obviously.
>>>>
>>>> When lyx starts up, it emits a complaint
>>>> 	QSettings: error creating /fs-sj1-15/darnstein/.qt
>>>>
>>>> When I try to save a lyx document to my home directory, lyx complains
>>>> when it presents its file browser dialog box. It says
>>>> 	Could not read directory /fs-sj1-15
>>>>
>>>> It appears that lyx is trying to access the root directory (/). It
>>>> does not seem to know how to interpret the Windows syntax "//."
>>> This is because lyx uses the boostfs library with BOOST_POSIX defined,
>>> so any path of the form //xxx/yyy is normalized to /xxx/yyy.
>>> I understand that //machine/path is a windowism, but I think that it
>>> should be allowed on cygwin. Can this be seen a boost bug?
>>>
>> Possibly.  But I expect boost folks would argue that using UNC syntax doesn't
>> fit with POSIX semantics.  Currently, the code keys off of BOOST_POSIX or
>> BOOST_WINDOWS defines to determine the API to use.  Under Cygwin, BOOST_POSIX
>> makes the most sense but with that definition comes the restriction of no UNC
>> paths.  Obviously, the workaround for now is to just mount the path in Cygwin
>> to create a POSIX path to use.  This will make Lyx happy without too much of a
>> burden for the user.
> 
> As I explained in an another mail, in this case the culprit is qt3
> and not boost. I suspected boost because of the following comment
> in boost/libs/filesystem/src/path_posix_windows.cpp:
> 
>   // POSIX & Windows cases: "", "/", "/foo", "foo", "foo/bar"
>   // Windows only cases: "c:", "c:/", "c:foo", "c:/foo",
>   //                     "prn:", "//share", "//share/", "//share/foo"
> 
> but, even if UNC paths are not taken into account, the leading '/' is
> not stripped out. Instead, it is qt3 that strips it and I also posted
> a patch allowing UNC paths in qt3. Maybe it can be applied to the cygwin
> package?
> 

Possibly.  That's a determination for the qt maintainer.  In the meantime,
my suggested workaround should still work.

-- 
Larry Hall                              http://www.rfk.com
RFK Partners, Inc.                      (508) 893-9779 - RFK Office
216 Dalton Rd.                          (508) 893-9889 - FAX
Holliston, MA 01746

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/