This patch replaces:
- {var}>... redirections with functionally identical eval construct +
explicit FDs
- ^^ and ,, case modifiers with temporary shopt
This allows us to lower minimum required bash version
to at least 3.1 (with current code).
Signed-off-by: Michal Soltys <soltys@ziu.info>
mostly with reference to earlier commit:
- bash doesn't need unsetting locals
- make normalize_path() a bit faster, also make sure we remove all
trailing slashes
- normalize paths before tests
Signed-off-by: Michal Soltys <soltys@ziu.info>
- IFS was not preserved, and modified value could leak to outside functions
- the '.' relative path should be returned for arguments such as /x/y/z
/x/y - but not for $1 == $2 ones
- $1 == $2 is self-looping link, so it returns final component of its
name
Signed-off-by: Michal Soltys <soltys@ziu.info>
Bash shell expands all ${parameter} before evaluating a command.
For multiple declarations and assignments within the same 'local' command,
then new variables or new values that appear towards the left
do not affect parameter expansion towards the right.
--
John Reiser, jreiser@BitWagon.com
>From 507ad6f66fc66f868a9e5fdd3806e012c4022baa Mon Sep 17 00:00:00 2001
From: John Reiser <jreiser@BitWagon.com>
Date: Fri, 23 Sep 2011 07:37:43 -0700
Subject: [PATCH] Parameter expansion occurs before command evaluation.
${parameter} on the right is expanded before evaluating "local var=value"
on the left.
Use bash "[[ string =~ pattern ]]" instead of "egrep -q".
Replace control-dominated serial fondling
for var in $(proc1); do proc2 var; done
with data-dominated parallel pipeline
proc1 | while read var; do proc2 var; done
Together this is a large savings.
[harald@redhat.com: fixed network kernel module filter]
The local variable _mpargs in function instmods() in file dracut-functions
looks peculiar. The documentation is non-existent, but still ...
First, $_mpargs is not passed to modprobe via for_each_kmod_dep.
This is strange because my guess is that "_mpargs" means
"extra arguments for modprobe".
Second, the leading "--" will be lopped when a leading pathname
is stripped via
_mod=${_mod##*/}
It seems to me that a leading "--" should inhibit modification.
Here's the corresponding patch to current HEAD (from dracut-013.)
This patch adds a new option --force-add, which
can force dracut to load some module when -H
is specified.
Signed-off-by: Amerigo Wang <amwang@redhat.com>
This update adds support for sort-of corner case - when explicitly
specified binary (e.g. through dracut_install or inst) is a library
itself.
In such case, we would expect the binary to undergo typical
library-related handling (symlinks and such).
Apart from that, the patch cleans indenting and a few unused variables
in inst_binary() (probably leftovers from the past ?)
Signed-off-by: Michal Soltys <soltys@ziu.info>
It's dash compatible to be used also at boot-time. For now it's included
by dracut-functions and replaces functions: dinfo(), dwarning() and
derror(). New options are introduced: -L|--stdlog, and -q|--quiet to
control stderr verbosity. Logging to file or syslog may be controlled by
options set in config file.
Note that code is not adjusted to the meaning of the new logging
functions, yet.
Doxygen formatted documentation (as a proposal, by the way) is included
in dracut-logger.
inst_dir used the following to try to resolve a relative path:
[[ $target = ${target##*/} ]] && target="${file%/*}/$target"
inst_dir $target
This will only match if $target has no slashes, so something like
/usr/bin -> ../sbin would result in: inst_dir ../sbin, or
/usr/share -> local/share would result in: inst_dir local/share
which is not going to do the right thing.
Instead, we resolve any non-absolute link, like so:
[[ $target == ${target#/} ]] && target=$(dirname "$file")/$target
Thus /usr/bin -> ../sbin results in: inst_dir /usr/../sbin, and
/usr/share -> local/share results in: inst_dir /usr/local/share
which is what you would expect.
Michal Soltys