Re: old_sigset_t vs sigset_t

Jason Flynn (flynnjs@mobility.bt.co.uk)
Tue, 09 Dec 1997 10:30:18 +0000


I've just gone from libc-5.4.33 -> 5.4.38 as this is what linuxhq
suggests
is needed for 2.1 - I'm (trying) to go 2.1 from today.

Kerneld 2.1.55 and mount 2.7g fail to compile due to sigset_t

C. Scott Ananian wrote:
>
> On Mon, 8 Dec 1997 "Steven N. Hirsch" <shirsch@ibm.net> wrote:
>
> > Any and all applications using signals now require editing to compile
> > against recent 2.1.x kernels. Is this really what was intended, or is
> > there a complimentary change which must occur in one's header files?
>
> This is not so. Check your header files. The kernel change should not
> show up in user-mode applications. There is/was a byte-endian thing that
> showed up in user-land in kernel 2.1.7[01?], but that was squashed.
> Somehow your user apps are using kernel headers.

Note the /usr/include/signal.h includes linux/signal.h

> > [..] I am running linux libc-5.4.33 [...]
>
> I'm running 5.4.38 with no problems.
>
> This list gets real silent when you report imaginary problems.
> --Scott
Ouch