On Sun, 29 Jan 2006, Garrett Rooney wrote:

> On 1/29/06, M Joonas Pihlaja <[EMAIL PROTECTED]> wrote:

> > SunOS mitzi 5.9 Generic_118558-09 sun4u sparc SUNW,Ultra-Enterprise

> > testflock hangs the process, and after vigorous killing, the
> > process becomes a zombie.

This is possibly some kind of strange NFS locking issue.  The
directory where I first tried this was NFS mounted.  Running
under dbx, I can see it blocking indefinitely in fcntl():

  [3] fcntl(0x3, 0x7, 0xffbff5ec, 0xff34f860, 0x0, 0xf4121), at 0xff25df28
=>[4] apr_file_lock(thefile = 0xf40d0, type = 2), line 46 in "flock.c"
  [5] test_withlock(tc = 0xffbff6f0, data = (nil)), line 68 in "testflock.c"

Unfortunately I can't quit from dbx except by killing it with
SIGTERM.  On the other hand, when compiling and running the tests
in /tmp, testflock completed successfully.

> > After removing the testflock case, testsockets segfaults.

In sendto_receivefrom(), the calls to apr_sockaddr_get_info()
fail with, essentially, EAI_NONAME from getaddrinfo(), and this
results in NULL 'from' and 'to' sockaddrs => segfault when
apr_socket_bind() is passed a NULL sockaddr.

Regards,

Joonas

Reply via email to