Description of problem:


Version-Release number of selected component (if applicable):
ksh-20100621-12.el6_2.1.x86_64

How reproducible:
Always

Steps to Reproduce:
1. su -
2. ksh
3. A=${pvs)
  
Actual results:
File descriptor 11 (/tmp/sf3e.p59 (deleted)) leaked on pvs invocation. Parent PID 22694: ksh


Expected results:
(no warnings)

Additional info:
This is not reproducible on RHEL5.7
So far I have only reproduced it using execution of the LVM command through a variable assignment as per the syntax above.

Further analysis of an strace shows the following ...

ksh on rhel57 (ksh-20100202-1.el5_6.6) does the following, dupping FD#10 to FD#1 and then closing FD#10 after a clone of pid 13008.

13007 fstat(1, {st_mode=S_IFSOCK|0200, st_size=0, ...}) = 0
13007 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x2b1db3dcefe0) = 13008
13007 fstat(5, {st_mode=S_IFSOCK|0400, st_size=0, ...}) = 0
13007 fstat(5, {st_mode=S_IFSOCK|0400, st_size=0, ...}) = 0
13007 close(1)                          = 0
13007 close(1)                          = -1 EBADF (Bad file descriptor)
13007 fcntl(10, F_DUPFD, 1)             = 1
13007 close(10)                         = 0

but ksh on rhel61 (ksh-20100621-6.el6.x86_64) does the following, dupping FD#1 to FD#11 then forking and dupping FD#11 to FD#1 and closing FD#11 after the fork.

10957 rt_sigprocmask(SIG_BLOCK, [HUP INT QUIT PIPE], [], 8) = 0
10957 vfork()                           = 10964
10957 rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
10964 rt_sigprocmask(SIG_SETMASK, [],  <unfinished ...>
10957 close(1 <unfinished ...>
10964 <... rt_sigprocmask resumed> NULL, 8) = 0
10957 <... close resumed> )             = 0
10957 fcntl(11, F_DUPFD, 1)             = 1
10964 execve("/sbin/lvs", ["lvs"], [/* 31 vars */] <unfinished ...>
10957 close(11)                         = 0

So it looks like because RHEL5.7 uses clone, we are able to close FD#10 after a clone because

       "Unlike fork(2), these calls allow the child process to share parts of its  execution
        context with the calling process, such  as  the  memory  space,  the table  of  file
        descriptors, and the table of signal handlers."

but because RHEL6.1 uses vfork we cannot close FD#11 after the fork.

So this looks like a bug introduced in the new version of ksh.