Memory leak in sp_open?

[stephenmathieson]

The example leaks 48 bytes (see this valgrind report):

#include <sophia.h>
#include <stdio.h>

int
main(void) {
  char *path = "/tmp/foodb";
  void *env = NULL;
  void *db = NULL;
  int rc = 0;

  if (!(env = sp_env())) {
    fprintf(stderr, "Failed to allocate environment\n");
    return 1;
  }

  rc = sp_ctl(env, SPDIR, SPO_CREAT|SPO_RDWR, path);
  if (-1 == rc) {
    sp_destroy(env);
    fprintf(stderr, "Failed to set database path\n");
    return 1;
  }

  rc = sp_ctl(env, SPGC, 1);
  if (-1 == rc) {
    sp_destroy(env);
    fprintf(stderr, "Failed to enable garbage collector\n");
    return 1;
  }

  if (!(db = sp_open(env))) {
    sp_destroy(env);
    fprintf(stderr, "Failed to open database\n");
    return 0;
  }

  printf("Opened database %s\n", path);

  sp_destroy(db);
  sp_destroy(env);
  return 0;
}

Is something wrong with the example, or is there a leak in Sophia?

[stephenmathieson]

bump -- any thoughts?

[pmwkaa]

~/craft/sophia/test valgrind ./sophia-leak
==11039== Memcheck, a memory error detector
==11039== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
==11039== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
==11039== Command: ./sophia-leak
==11039==
Opened database /tmp/foodb
==11039==
==11039== HEAP SUMMARY:
==11039== in use at exit: 0 bytes in 0 blocks
==11039== total heap usage: 17 allocs, 17 frees, 67,160 bytes allocated
==11039==
==11039== All heap blocks were freed -- no leaks are possible
==11039==
==11039== For counts of detected and suppressed errors, rerun with: -v
==11039== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

~/craft/sophia/test valgrind --version
valgrind-3.7.0

That is funny, I might have outdated valgrind, i will check it up :)

[stephenmathieson]

Very strange -- I've got the same version:

$ valgrind --version
valgrind-3.7.0

[pmwkaa]

There might be difference in pthread implementation (libc version), i need to checkout proper condition variable cleanup on exit. Thanks)

[stephenmathieson]

Do you get the same successful results when running valgrind --leak-check=full ./sophia-leak?

[pmwkaa]

yeah, just rechecked. As far as i can see in source code, there are missed condition variable destruction on sp_destroy. Seems to be pthread implementation started to allocate memory for their own purposes, so the problem came up :)

[stephenmathieson]

Interesting. If you point me in the right direction, I can try my hand at a fix :)