(sorry for the previous garbled message, hopefully this one comes out
better)Configuration Information [Automatically generated, do not
change]:Machine: x86_64
OS: linux-gnu
Compiler: gcc
Compilation CFLAGS: -DPROGRAM='bash' -DCONF_HOSTTYPE='x86_64'
-DCONF_OSTYPE='linux-gnu' -DCONF_MACHTYPE='x86_64-pc-linux-gnu'
-DCONF_VENDOR='pc' -DLOCALEDIR='/usr/share/locale' -DPACKAGE='bash' -DSHELL
-DHAVE_CONFIG_H -I. -I../bash -I../bash/include -I../bash/lib -g -O2 -Wall
uname output: Linux jpadesk2.mtv.corp.google.com 2.6.32-gg252-generic
#gg252-Ubuntu SMP Mon Sep 13 22:09:00 UTC 2010 x86_64 GNU/Linux
Machine Type: x86_64-pc-linux-gnu
Bash Version: 4.1
Patch Level: 5
Release Status: release
Description:
Named pipes are being leaked.
The named pipe created for the "echo stuff" in
while read x; do echo $x ; done < <(echo stuff)
is never closed.
Repeat-By:
Running the following script.
<test_named_pipe_leak.bash4.sh>
#!/bin/bash
func1() {
local arr
while read arr; do
# The following cmd | cmd causes the leaked named pipe to be cleaned up.
[ "$do_workaround" == yes ] && exec | exec
# Normally code here would modify some env var. So can't use
# "cmd | while read...".
done< <(echo hello)
echo func1_lsof=$(lsof | grep -c test_name.*pipe)
}
main() {
local do_workaround=$1
local tmpworkdir=/tmp/test_named_pipe_leak.tmp.$$
mkdir -p $tmpworkdir
lsof -p $$ > $tmpworkdir/lsof_start
for ((i=0; i < 10; i++)); do
func1
done
lsof -p $$ > $tmpworkdir/lsof_end
diff -us $tmpworkdir/lsof_start $tmpworkdir/lsof_end
rm -rf $tmpworkdir
}
main "$@"
</test_named_pipe_leak.bash4.sh>
Running
~/tmp/test_named_pipe_leak.bash4.sh no
outputs
<output_txt>
func1_lsof=4
func1_lsof=5
func1_lsof=6
func1_lsof=7
func1_lsof=8
func1_lsof=9
func1_lsof=10
func1_lsof=11
func1_lsof=12
func1_lsof=13
--- /home/jpa/tmp/testdir.26477/lsof_start 2010-11-17 13:12:07.904919000
-0800
+++ /home/jpa/tmp/testdir.26477/lsof_end 2010-11-17 13:12:11.280668000
-0800
@@ -22,4 +22,14 @@
test_name 26477 jpa 0u CHR 136,0 0t0 3 /dev/pts/0
test_name 26477 jpa 1u CHR 136,0 0t0 3 /dev/pts/0
test_name 26477 jpa 2u CHR 136,0 0t0 3 /dev/pts/0
+test_name 26477 jpa 54r FIFO 0,8 0t0 5655855 pipe
+test_name 26477 jpa 55r FIFO 0,8 0t0 5655740 pipe
+test_name 26477 jpa 56r FIFO 0,8 0t0 5655627 pipe
+test_name 26477 jpa 57r FIFO 0,8 0t0 5655514 pipe
+test_name 26477 jpa 58r FIFO 0,8 0t0 5655401 pipe
+test_name 26477 jpa 59r FIFO 0,8 0t0 5655288 pipe
+test_name 26477 jpa 60r FIFO 0,8 0t0 5655175 pipe
+test_name 26477 jpa 61r FIFO 0,8 0t0 5655051 pipe
+test_name 26477 jpa 62r FIFO 0,8 0t0 5654938 pipe
+test_name 26477 jpa 63r FIFO 0,8 0t0 5654823 pipe
test_name 26477 jpa 255r REG 0,27 712 6002620
/home/jpa/tmp/test_named_pipe_leak.bash4.sh (mtvhome30.nfs:/vol/mtvhome30/jpa)
</output_txt>
Running
~/tmp/test_named_pipe_leak.bash4.sh yes
outputs
<output_txt>
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
func1_lsof=3
Files /home/jpa/tmp/testdir.26915/lsof_start and
/home/jpa/tmp/testdir.26915/lsof_end are identical
</output_txt>
