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-redhat-linux-gnu' -DCONF_VENDOR='redhat' -DLOCALEDIR='/usr/share/locale' -DPACKAGE='bash' -DSHELL -DHAVE_CONFIG_H -I. -I. -I./include -I./lib -D_GNU_SOURCE -DRECYCLES_PIDS -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic uname output: Linux office 2.6.34.7-61.fc13.x86_64 #1 SMP Tue Oct 19 04:06:30 UTC 2010 x86_64 x86_64 x86_64 GNU/Linux Machine Type: x86_64-redhat-linux-gnu
Bash Version: 4.1 Patch Level: 7 Release Status: release Description: Bash does not behave well when under artificial fd pressure due to ulimit -n. It issues a spurious warning to stderr because it tries to save necessary fds starting at 10. Compare this with ksh93, which saves fds starting at 3. Many other shells (for example, dash or BSD /bin/sh) exit with non-zero status if they can't use fd 10, rather than proceeding onwards. At any rate, bash MUST exit with failure if it cannot save an fd, even if you decide that it is unsafe to copy ksh93's action of saving at fd 3 rather than 10. The current behavior results in bash silently corrupting stdin! Merely warning me that you lost stdin, but still exiting with 0 status, is a disservice. Repeat-By: $ ksh -c 'ulimit -n 10; : <&-; exec 3<&0'; echo $? 0 $ ksh -c 'ulimit -n 3; : <&-; exec 3<&0'; echo $? ksh[1]: open file limit exceeded [Invalid argument] 1 $ ksh -c 'ulimit -n 10; : <&-'; echo $? 0 $ bash -c 'ulimit -n 11; : <&-'; echo $? 0 $ # This next step should either silently work or die with an error $ bash -c 'ulimit -n 10; : <&-'; echo $? bash: redirection error: cannot duplicate fd: Invalid argument 0 $ # Proof that you silently lost stdin during the : <&- step. $ bash -c 'ulimit -n 10; : <&-; exec 3<&0'; echo $? bash: redirection error: cannot duplicate fd: Invalid argument bash: 0: Bad file descriptor 1 Fix: The bare minimum fix is to error out if you cannot dup() a fd to preserve it across the duration of a temporary redirection. But the nicer fix would be to follow ksh's lead and start saving fd's at 3, even if this means doing more bookkeeping of which fds have been put into use by the user vs. used by bash (that is, if stdin was saved by bash in fd 3, but then the script asks to redirect fd 3, then bash would have to re-save fd 3 to somewhere else). -- Eric Blake ebl...@redhat.com +1-801-349-2682 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
