Re: Microsoft SFU

[Bob Proulx]

Dr. Franz Fehringer wrote:
> I tried yesterday to install the current coreutils version (5.3.0
> IIRC) on Microsoft ServicesForUnix (SFU) 3.5.

Thanks for the report.  However I can't say as I have ever even heard
of that particular flavor of platform before.  (My first reaction was,
"What does 'uname -a' say?  Oh."  My second reaction was, "What
version of GCC and mostly importantly what version of GLIBC are you
using?  Oh.")  I don't think we on the GNU coreutils mailing list are
going to be able to help you very much.

Since you are working on a MS platform, you should check with the
cygwin folks.  They have already ported coreutils to MS-Windows.  It
sounds like you are trying do the same thing.  I am sure there is a
lot of learning and experience that can be shared from them.  They
also provide precompiled binaries.  Would that be sufficient for you?

  http://cygwin.com/lists.html

Having said the above, here is my best shot at your problem.

> The configure script halted with an error, because it could not find
> out how to enumerate the list of mounted filesystems.

This sounds like a configure script issue of some sort.  If you look
in the generated config.log file there should be some clues as to what
was being tested and what was happening when configure bailed out.  If
it was a script test or if it was trying to compile a C test program
or whatever.  Crosscheck the log in config.log with the actual test
being run in the configure script.  By looking at the test it was
performing and recreating it yourself you should be able to recreate
the problem in a smaller test case.

For example, you will see many stanzas in the configure script where
it creates a conftest.c file and tries to compile it.  Extract the
conftest.c file yourself from the configure script and compile it
using the same compile line that configure is using.  At that time you
will see the same failure that happens when configure is run.  That
information is the important information for improving configure.

> But there must be a way, since SFU contains a working df command.

MS is so quite different from GNU/UNIX systems in general that even if
a command line program 'df' exists it does not mean that the program
implementing it resembles anything similar.  You can't tell from the
outside.  So while it may seem similar on the surface it is just as
likely that the implementation may be so different that it is not
practical to try to share the implementation.

Remember that the whole point of GNU software is to implement the GNU
operating system.  Ports to other systems, even non-free systems, are
great because it allows people there to see the value of free
software.  But if there is a conflict in implementation then it is
obvious which way it must be decided.

Check with the Cygwin folks and see what information you can find
there.

Bob