Re: [Jfs-discussion] mkfs.jfs options

[Dave K. <sh...@au...>]

On Fri, 2004-04-30 at 09:23, Peter Weilbacher wrote:
> On Fri, 30 Apr 2004, Dave Kleikamp wrote:
> 
> > I don't think the -O option would affect the fcntl or flock functions.
> > The case-insensitivity is the big factor with the -O flag.  I don't know
> > anything about emerge, but if it is trying to use file names that only
> > differ by case, such as fcntl.py and FCNTL.py, they will refer to the
> > same file.  I suspect that this may be causing the problem.
> 
> OK, this makes sense, I should have noticed that myself. Although I
> don't know much about the workings of Python, there are FCNTL.py and
> fnctl.so in /usr/lib/python2.3/, probably those are confused when
> queried case-insensitively.

I'd bet if you check on a sane case-sensitive filesystem, you'd also
find a fctnl.py and/or FCNTL.so.  On a case-insensitive file system,
there can only be one of each.

> But if "-O" does nothing else why am I not able to get _any_ access to
> the JFS partition from OS/2 when formatted without the "-O" switch?  A
> case-sensitive FS should only be a problem for running programs under
> OS/2, because e.g. DLLs are always referenced in upper case. For
> datafiles it should not matter, or am I wrong there?

The problem is that the internal directory structures are either ordered
case-sensitive (default linux) or case-insensitive (OS/2).  The Linux
code knows how to traverse both, but the OS/2 code only understands the
case-insensitive structures.

> Anyway, thanks for solving that mystery!
> 
> Cheers,
>    Peter.
-- 
David Kleikamp
IBM Linux Technology Center