[PATCH v2 00/25] Simplify inclusion conditionals

[pubg lover]

Hindi language please

On Mon, Jun 20, 2022, 8:02 PM Andrew Clayton <andrew at digital-domain.net>
wrote:

> On Mon, 20 Jun 2022 13:36:00 +0100
> Andrew Clayton <andrew at digital-domain.net> wrote:
>
> > On Mon, 20 Jun 2022 13:28:41 +0100
> > Andrew Clayton <andrew at digital-domain.net> wrote:
> >
> > > But then I did a make clean and ./configure && make and all was good.
> >
> > Hmm, so taking my own advice, _now_ I see a load of errors... will
> > investigate.
>
> Not got right to the bottom of it, but here's where we're at.
>
> The first error is
>
> cc -c -pipe -fPIC -fvisibility=hidden -O -W -Wall -Wextra
> -Wno-unused-parameter -Wwrite-strings -Wmissing-prototypes -Werror -g   -I
> src -I build   \
>                       \
>                      \
> -o build/src/nxt_socketpair.o \
> -MMD -MF build/src/nxt_socketpair.dep -MT build/src/nxt_socketpair.o \
> src/nxt_socketpair.c
> In file included from /usr/include/sys/socket.h:33,
>                  from /usr/include/netinet/in.h:23,
>                  from /usr/include/netdb.h:27,
>                  from src/nxt_unix.h:157,
>                  from src/nxt_main.h:31,
>                  from src/nxt_socketpair.c:7:
> src/nxt_socket_msg.h:32:54: error: invalid application of ‘sizeof’ to
> incomplete type ‘nxt_socket_cred_t’ {aka ‘struct ucred’}
>    32 |     (CMSG_SPACE(2 * sizeof(int)) +
> CMSG_SPACE(sizeof(nxt_socket_cred_t)))
>       |
> ^~~~~~~~~~~~~~~~~
>
> On Linux/glibc at least, struct ucred is defined in
> /usr/include/bits/socket.h as
>
> #ifdef __USE_GNU
>
> /* User visible structure for SCM_CREDENTIALS message */
>
> struct ucred
>
> {
>
>   pid_t pid;                    /* PID of sending process.  */
>
>   uid_t uid;                    /* UID of sending process.  */
>
>   gid_t gid;                    /* GID of sending process.  */
>
> };
>
> #endif
>
> So it needs _GNU_SOURCE defined, as confirmed by this test case
>
> #define _GNU_SOURCE 1
>
>
> #include <sys/socket.h>
>
>
> int main(void)
>
> {
>
>         sizeof(struct ucred);
>
>         return 0;
>
> }
>
> Commenting out the #define makes it fail to compile.
>
> $ make m
> cc     m.c   -o m
> m.c: In function ‘main’:
> m.c:9:16: error: invalid application of ‘sizeof’ to incomplete type
> ‘struct ucred’
>     9 |         sizeof(struct ucred);
>       |                ^~~~~~
> make: *** [<builtin>: m] Error 1
>
> Now previously malloc.h was being included _after_ _GNU_SOURCE was being
> defined.
> malloc.h seems to somehow be interfering with that, as demonstrated by
>
> #include <malloc.h>
>
>
> #define _GNU_SOURCE 1
>
>
> #include <sys/socket.h>
>
>
> int main(void)
>
> {
>
>         sizeof(struct ucred);
>
>         return 0;
>
> }
>
>
> which fails to build.
>
> Indeed if in unit, I comment out the malloc.h include, it builds, or if
> I define _GNU_SOURCE before it, it builds.
>
> Hmm, actually, including any system (at least) header file before
> _GNU_SOURCE is defined breaks the test program.
>
> Arghhhh....
>
> Of course, all these header files include features.h which deals with
> all the standards macros.
>
> So if we include a header file before _GNU_SOURCE is defined, then
> _FEATURES_H is defined, then if we define _GNU_SOURCE and include more
> header files, then features.h won't do anything due to _FEATURES_H
> already being defined and _GNU_SOURCE will have no effect as in
> /usr/include/features.h
>
> #ifdef  _GNU_SOURCE
> # define __USE_GNU      1
> #endif
>
> won't happen. (__USE_GNU is what glibc seems to use internally for
> _GNU_SOURCE).
>
> OK, so not every header file includes features.h, for example
> sys/syscall.h which is why things worked before.
>
> So maybe we did get to the bottom of it...
>
> So it just remains to see how to solve this current issue. But
> basically you need to define _GNU_SOURCE before including any header
> file that includes features.h.
>
> Cheers,
> Andrew
>
> _______________________________________________
> unit mailing list -- unit at nginx.org
> To unsubscribe send an email to unit-leave at nginx.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nginx.org/pipermail/unit/attachments/20220620/774528dd/attachment.htm>