Using __ppoll64 internally

[Cristian Rodríguez]

On Thu, Sep 9, 2021 at 4:18 PM Florian Weimer <fweimer@redhat.com> wrote:

> So you need to provide more details about your build environment, and
> more of the warning message.  There should be a location of the use of
> the macro.

I will exemplify then with the problematic

diff --git a/resolv/res_send.c b/resolv/res_send.c
index 9f86f5fe47..ce612bd55c 100644
--- a/resolv/res_send.c
+++ b/resolv/res_send.c
@@ -927,9 +926,8 @@ send_dg(res_state statp,
 {
        const HEADER *hp = (HEADER *) buf;
        const HEADER *hp2 = (HEADER *) buf2;
-       struct timespec now, timeout, finish;
+       struct __timespec64 now, timeout, finish;
        struct pollfd pfd[1];
-       int ptimeout;
        struct sockaddr_in6 from;
        int resplen = 0;
        int n;
@@ -979,14 +977,11 @@ send_dg(res_state statp,
                evSubTime(&timeout, &finish, &now);
                need_recompute = 0;
        }
-       /* Convert struct timespec in milliseconds.  */
-       ptimeout = timeout.tv_sec * 1000 + timeout.tv_nsec / 1000000;
-
        n = 0;
        if (nwritten == 0)
          n = __poll (pfd, 1, 0);
        if (__glibc_unlikely (n == 0))       {
-               n = __poll (pfd, 1, ptimeout);
+               n = __ppoll64 (pfd, 1, &timeout, NULL);
                need_recompute = 1;
        }
        if (n == 0) {

Here using __ppoll64 should be valid, as it is the correct function
that takes an struct __timespec64 as timeout parameter.
It is indeed a valid callable function but it is #define'd in the
headers to a symbol that is nowhere to be declared.