timer_create and timer_delete cause valgrind errors

[Mark P]

Using timer_create/timer_delete causes an untold number of valgrind errors in programs which use timers heavily.

glibc 2.9-4
Linux 2.6.28-15-generic #52-Ubuntu SMP Wed Sep 9 10:49:34 UTC 2009 i686 GNU/Linux
Ubuntu 9.04

Some of the errors may be attributable to this bug:
https://bugs.launchpad.net/bugs/264868

However, the following is the minimal program for generating valgrind errors using timer_create/delete and doesn't even start the timer so is not attributable to the above bug. The error occurs whether the timer is started or not.

#include <time.h>
#include <signal.h>
#include <memory.h>
#include <assert.h>

void callback( sigval_t parameter )
{
}

int main( int argc, char** argv )
{
    sigevent_t se;
    memset( &se, 0 , sizeof(se) ); // just to be sure all of se is initialised
    se.sigev_notify = SIGEV_THREAD;
    se.sigev_value.sival_ptr = 0;
    se.sigev_notify_function = callback;
    se.sigev_notify_attributes = 0;

    timer_t t;
    int err = timer_create( CLOCK_MONOTONIC, &se, &t );
    assert( err != -1 );
    timer_delete( t );

    return 0;
}

When compiled using 'gcc -g main.c -lrt' this will produce the following valgrind output:

$ valgrind --track-origins=yes --leak-check=full ./a.out
==11225== Memcheck, a memory error detector.
==11225== Copyright (C) 2002-2008, and GNU GPL'd, by Julian Seward et al.
==11225== Using LibVEX rev 1884, a library for dynamic binary translation.
==11225== Copyright (C) 2004-2008, and GNU GPL'd, by OpenWorks LLP.
==11225== Using valgrind-3.4.1-Debian, a dynamic binary instrumentation framework.
==11225== Copyright (C) 2000-2008, and GNU GPL'd, by Julian Seward et al.
==11225== For more details, rerun with: -v
==11225==
==11225== Conditional jump or move depends on uninitialised value(s)
==11225== at 0x40454D4: timer_delete (in /lib/tls/i686/cmov/librt-2.9.so)
==11225== by 0x8048527: main (main.c:22)
==11225== Uninitialised value was created by a heap allocation
==11225== at 0x4026FDE: malloc (vg_replace_malloc.c:207)
==11225== by 0x404534E: timer_create (in /lib/tls/i686/cmov/librt-2.9.so)
==11225== by 0x80484EF: main (main.c:20)
==11225==
==11225== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 15 from 1)
==11225== malloc/free: in use at exit: 136 bytes in 1 blocks.
==11225== malloc/free: 2 allocs, 1 frees, 192 bytes allocated.
==11225== For counts of detected errors, rerun with: -v
==11225== searching for pointers to 1 not-freed blocks.
==11225== checked 84,192 bytes.
==11225==
==11225==
==11225== 136 bytes in 1 blocks are possibly lost in loss record 1 of 1
==11225== at 0x4025092: calloc (vg_replace_malloc.c:397)
==11225== by 0x401134B: _dl_allocate_tls (in /lib/ld-2.9.so)
==11225== by 0x41B3672: pthread_create@@GLIBC_2.1 (in /lib/tls/i686/cmov/libpthread-2.9.so)
==11225== by 0x4046501: (within /lib/tls/i686/cmov/librt-2.9.so)
==11225== by 0x41B95CF: pthread_once (in /lib/tls/i686/cmov/libpthread-2.9.so)
==11225== by 0x80484EF: main (main.c:20)
==11225==
==11225== LEAK SUMMARY:
==11225== definitely lost: 0 bytes in 0 blocks.
==11225== possibly lost: 136 bytes in 1 blocks.
==11225== still reachable: 0 bytes in 0 blocks.
==11225== suppressed: 0 bytes in 0 blocks.

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Nate Eldredge (nate-thatsmathematics)]

Cannot reproduce under 19.10 amd64.