[arch-projects] [initscripts] [PATCH] Fix: don't use install where target file can't/shouldn't be removed

[Kurt J. Bosch]

Seblu, 2011-06-27 00:51:
> On Mon, Jun 27, 2011 at 12:44 AM, Kurt J. Bosch
> <kjb-temp-2009 at alpenjodel.de>  wrote:
>> Seblu, 2011-06-27 00:28:
>>>
>>> On Sun, Jun 26, 2011 at 11:51 PM, Kurt J. Bosch
>>> <kjb-temp-2009 at alpenjodel.de>    wrote:
>>>>
>>>> ---
>>>>   functions  |    7 +++++++
>>>>   rc.sysinit |    4 ++--
>>>>   2 files changed, 9 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/functions b/functions
>>>> index 43414c0..9ad75a9 100644
>>>> --- a/functions
>>>> +++ b/functions
>>>> @@ -428,6 +428,13 @@ install_file() {
>>>>         install -T "$@"<(printf '%s' "$content") "$file"
>>>>   }
>>>>
>>>> +# install would fail because files can't be removed from /proc
>>>> +# moreover we don't want to remove nodes from /dev
>>>> +write_file() {
>>>> +       local file=$1 content=$2
>>>> +       printf '%s' "$content">| "$file"
>>>> +}
>>>> +
>>>>   bootlogd_stop() {
>>>>         [[ -f /run/bootlogd.pid ]] || return 0
>>>>         touch /var/log/boot
>>>> diff --git a/rc.sysinit b/rc.sysinit
>>>> index dcc88f5..2c08f02 100755
>>>> --- a/rc.sysinit
>>>> +++ b/rc.sysinit
>>>> @@ -209,7 +209,7 @@ status "Activating Swap" swapon -a
>>>>   RANDOM_SEED=/var/lib/misc/random-seed
>>>>   [[ -f $RANDOM_SEED ]]&&
>>>>         status "Initializing Random Seed" \
>>>> -               install_file /dev/urandom "$(<    "$RANDOM_SEED" )"
>>>> +               cp "$RANDOM_SEED" /dev/urandom
>>>>
>>>>   stat_busy "Removing Leftover Files"
>>>>         rm -rf /etc/{nologin,shutdownpid} /forcefsck /tmp/*
>>>> /tmp/.*&>/dev/null
>>>> @@ -225,7 +225,7 @@ stat_done
>>>>
>>>>   [[ $HOSTNAME ]]&&
>>>>         status "Setting Hostname: $HOSTNAME" \
>>>> -               install_file /proc/sys/kernel/hostname "$HOSTNAME"
>>>> +               write_file /proc/sys/kernel/hostname "$HOSTNAME"
>>>>
>>>
>>> Do you call a function to do the job of a redirection?
>>>
>> Yep. We need to do it _within_ status.
>
> Reading (or not) your commit message I thought you wanted to do that.
> What do you think about doing you eval inside status?
>
Would be rather difficult because eval does word splitting.

> Or better, just rollback from your previous patch where you assert
> this write can be done in one line with status instead of a classic
> status_start / status_done?
> In this case, the use of status is not the best choice I think.
>
Using status + write_file is better in this case IMHO for some reasons:
* It provides DONE/FAIL handling.
* We avoid the common pitfall of changing the from stat_busy/stat_done 
to the non working status with redirection later as happened for 
"Initializing Random Seed" two moths ago without beeing fixed until now.
http://projects.archlinux.org/initscripts.git/commit/?id=1de599669a643f82ec0f0c7870335620e74b0765
(Would be fixed by "[PATCH] rc.sysinit: Fix Initializing Random Seed")

-- 
Kurt