stime() is no longer declared in time.h for rawhide

[Martin Gansser]

was the stime () declaration moved in another header file ?
What is the reason for this.

Regards
Martin

[Tom Hughes]

On 13/12/2019 07:46, Martin Gansser wrote:

>  was the stime () declaration moved in another header file ?
>  What is the reason for this. 
Commit history explains:

   https://sourceware.org/git/?p=glibc.git;a=commit;h=12cbde1da

But summary seems to be, as was said when the same question
was asked yesterday, to use clock_settime instead.

Tom

-- 
Tom Hughes (tom(a)compton.nu)
http://compton.nu/

[Martin Gansser]

ok, that means

stime () has been deprecated in glibc 2.31 and replaced with clock_settime ().

[Frank R Dana Jr.]

Yup. To be precise, quoting from the NEWS entry added for the change:

* The obsolete function stime is no longer available to newly linked
  binaries and it has been removed from <time.h> header.  This function
  has been deprecated in favor of clock_settime.

[Martin Gansser]

solved now with this patch:

diff --git a/eit.c b/eit.c
index 50d8229..82294dc 100644
--- a/eit.c
+++ b/eit.c
@@ -391,7 +391,9 @@ cTDT::cTDT(const u_char *Data)
   if (abs(diff) > MAX_TIME_DIFF) {
      mutex.Lock();
      if (abs(diff) > MAX_ADJ_DIFF) {
-        if (stime(&dvbtim) == 0)
+        timespec ts = {0};
+        ts.tv_sec = dvbtim;
+        if (clock_settime(CLOCK_REALTIME, &ts) == 0)
            isyslog("system time changed from %s (%ld) to %s (%ld)",
*TimeToString(loctim), loctim, *TimeToString(dvbtim), dvbtim);
         else
            esyslog("ERROR while setting system time: %m");

[Kevin Kofler]

Tom Hughes wrote:
>  Commit history explains:
>  
>     https://sourceware.org/git/?p=glibc.git;a=commit;h=12cbde1da 
For what it's worth, I think this is a completely pointless incompatibility.

It costs almost nothing to maintain this function (and in fact they have to 
do it anyway for old binaries), so I do not see the point of hiding it from 
the linker at all. All it means is that the applications have to copy&paste 
the same boilerplate everywhere (and do it wrong, e.g., Martin Gansser's 
patch will trigger an – admittedly harmless – compiler warning due to 
partial structure initialization). Sure, the new API is better because it 
supports split seconds, but if you do not have any nanoseconds to pass to 
it, there is no benefit in setting up the structure in the caller instead of 
letting glibc do it.

So I do not understand at all why glibc is now hiding this function from the 
compiler and the linker.

        Kevin Kofler