Hi Thomas, On 20/01/2026 15:10, Thomas Weißschuh wrote: > Interleaving inclusions of UAPI headers and libc headers is problematic. > Both sets of headers define conflicting symbols. To enable their > coexistence a compatibility-mechanism is in place. > > An upcoming change will define 'struct sockaddr' from linux/socket.h. > However sys/socket.h from libc does not yet handle this case and a > symbol conflict will arise. > > Furthermore libc-compat.h evaluates the state of the libc > inclusions only once, at the point it is included first. If another > problematic header from libc is included later, symbol conflicts arise. > This will trigger other duplicate definitions when linux/libc-compat.h > is added to linux/socket.h > > Move the inclusion of UAPI headers after the inclusion of the glibc > ones, so the libc-compat.h continues to work correctly.
Thank you for looking at this! Here is my (late, sorry) review for the modifications related to MPTCP: > diff --git a/tools/testing/selftests/net/mptcp/mptcp_diag.c > b/tools/testing/selftests/net/mptcp/mptcp_diag.c > index 8e0b1b8d84b6..af25ebfd2915 100644 > --- a/tools/testing/selftests/net/mptcp/mptcp_diag.c > +++ b/tools/testing/selftests/net/mptcp/mptcp_diag.c > @@ -1,11 +1,6 @@ > // SPDX-License-Identifier: GPL-2.0 > /* Copyright (c) 2025, Kylin Software */ > > -#include <linux/sock_diag.h> > -#include <linux/rtnetlink.h> > -#include <linux/inet_diag.h> > -#include <linux/netlink.h> > -#include <linux/compiler.h> > #include <sys/socket.h> > #include <netinet/in.h> > #include <linux/tcp.h> There is a remaining one (linux/tcp.h) here that you might want to move below too. > @@ -17,6 +12,12 @@ > #include <errno.h> > #include <stdio.h> > > +#include <linux/sock_diag.h> > +#include <linux/rtnetlink.h> > +#include <linux/inet_diag.h> > +#include <linux/netlink.h> > +#include <linux/compiler.h> Note that I just noticed this is the only file from this directory where the "includes" are not sorted by type and alphabetical order, see pm_nl_ctl.c as an example. A bit of a detail, but if you plan to send a v2, do you mind doing that too here while at it, please? If not, I can look at that later, but better to avoid doing that in parallel. Cheers, Matt -- Sponsored by the NGI0 Core fund.
