Re: running legacy apps compiled for 7.2 & 7.3 in x86-64

[Calvin Austin <calvin.austin@xxxxxxx>]

"Mike A. Harris" wrote:

> On Mon, 5 May 2003, Calvin Austin wrote:
>
> >So looking through the archives, has anyone cleanly added the ia32 libraries
> >back in. A glob on rh9 i386 packages didn't quite do the trick for me, all
> >sorts of dependencies and postinstall failures to unravel.
>
> A glob may be insufficient.  A proper installer would be nice to
> have, however for now, proper manual dependancy resolution may be
> necessary.  I'm sure there are probably some problems though in
> either case.  They're worth reporting in bugzilla though, or
> bringing up here at least.

hmm,


>
>
> >With Suse we experienced issues with floating point in 32bit emulation mode
> >amongst other things. (running 32bit in 64bit os mode)
>
> It is quite important to realize, that running 32bit binaries in
> compatibility submode of long mode under a 64bit OS, that you are
> not "emulating" 32bit code.  The 32bit code is running full steam
> ahead on the raw processor in it's 32bit compatibility mode.
> There is no emulation occuring.
>
> That said, I'm not aware of FPU related issues, but that is
> something we'd definitely want to hear more about.  Do you have
> more details about this?

The control word was being reset in the kernel when handling a signal. I haven't
looked at the redhat implementation but when running in 64bit long mode the
control word size is different depending on whether the app is 32bit or 64bit.
We ran some fp tests to find it but until I get a 32bit jvm running I can't run
the test set.

regards
calvin




>
>
> --
> Mike A. Harris     ftp://people.redhat.com/mharris
> OS Systems Engineer - XFree86 maintainer - Red Hat
>
> --
> AMD64-list mailing list
> AMD64-list@xxxxxxxxxx
> https://listman.redhat.com/mailman/listinfo/amd64-list