Re: [PATCH] Assume that pipe2 is always available

[Florian Weimer <fweimer at redhat dot com>]

On 04/17/2017 08:17 PM, Adhemerval Zanella wrote:
> On 13/04/2017 14:54, Florian Weimer wrote:
> > On 04/13/2017 07:39 PM, Adhemerval Zanella wrote:
> > > On 13/04/2017 12:37, Florian Weimer wrote:
> > > > The Debian patches (which are already required to build glibc before
> > > > this commit) contain an implementation of pipe2.
> > >
> > > I did not follow, which 'Debian patches' are you referring here?
> >
> > Upstream master contains an incomplete implementation of O_CLOEXEC support for Hurd.  For example, the file sysdeps/mach/hurd/accept4.c refers to the sock_to_o_flags identifier, but neither glibc, gnumach, nor hurd contain a definition.  The definition is found in a patch in the Debian package.  There is another patch which contains an implementation of pipe2:
> >
> > https://sources.debian.net/src/glibc/2.24-8/debian/patches/hurd-i386/tg-pipe2.diff/
> >
> > Anyone who wants to build glibc for Hurd needs those Debian patches, so I see no problem with applying this cleanup to upstream master.
> >
> > (From the Hurd perspective.  NaCl does not support pipe2, either.)
>
> Right, I think we need to sync on master since it seems required to actually
> build hurd.

Would you please clarify what you mean?  Do you suggest we have to pull 
in the Debian patches before we should apply this pipe2 cleanup?

Thanks,
Florian