Peter J. Stieber wrote:
PS = Peter J. Stieber
PS>> I recently posted mentioning that I was getting out of
PS>> memory (oom) errors with FC4 kernels 1656 and 1831 on
PS>> both x86_64 and i686 version of the kernel.
PS>>
PS>> http://www.redhat.com/archives/fedora-list/2006-February/msg01112.html
PS>>
PS>> I have simple Bash shell scripts (first line has #!/bin/bash)
PS>> that seem to start up many unicode_start processes, but it
PS>> doesn't occur every time I run the scripts. I haven't been
PS>> able to figure out what triggers this behavior.
RN = Robert Nichols
RN> Bug #181908 seems to be the same problem from a different
RN> trigger. See
RN> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=181809
RN> and add any comments you think are appropriate.
Thanks for pointing this out Robert. I was beginning to this I was
loosing my mind ;-)
I swear I searched redhat bugzilla for unicode_start and came up dry,
but thanks to your post I see 181908, 172059, and 179949.
And the final result of all this is that setting BASH_ENV=~/.bashrc
is now a Very Bad Thing(tm). Some of the files that get executed
when /etc/bashrc runs are themselves bash scripts. BASH_ENV gets
used by these non-interactive scripts, and if that leads back to
/etc/bashrc again the whole thing forms an infinite loop. Anything
in your .bashrc file that you need for non-interactive scripts
should be put in a separate file, and make _that_ the BASH_ENV
file. You can source the BASH_ENV file from your .bashrc if you
don't want stuff duplicated in the two files.
An alternative in .bashrc is to test whether "$-" includes the
letter "i" and only source /etc/bashrc if the shell is interactive.
--
Bob Nichols Yes, "NOSPAM" is really part of my email address.
