set_thread_buffer error

[Ove Kaaven]

On Mon, 17 Sep 2001 lawson_whitney at juno.com wrote:

> On 17 Sep 2001, Peter C wrote:
> 
> > No, it's standard Zipslack (the site doesn't give version numbers).
> >
> > I've looked all over Slackware's site and there's nothing mentioning
> > bugs with 64-bit support or a patch for it. I've got glibc 2.2.3 at
> > the moment, do you think upgrading to the latest version of glibc
> > (2.2.4) will help?
> 
> I don't think so.  That is starting to look to me like a umsdos or msdos
> bug, but to be sure I will have to compile a kernel with umsdos support,
> and I may as well patch to 2.4.9 while I am at it.

It's probably also a known bug; as far as I remember, Alan Cox refused to
apply a particular ftruncate() patch (which allows to grow files with
ftruncate() and expect the new parts to be filled with zeroes) to Linux's
FAT filesystem code, because he considered such behaviour broken. Alas,
Wine depends on this behaviour for shared memory sections (like those used
to communicate with the wineserver). (The same issue used to cause
problems with reiserfs too, but reiserfs fixed it.) Unless Alan changes
his mind, you'll probably need to mount some other filesystem at /tmp...