| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When functions provided by /etc/rc.d/functions is used, escape
sequence is output under systemd service units and logged into
/var/log/messages as follows:
# LANG=C systemctl cat foobar.service
# /etc/systemd/system/foobar.service
[Unit]
Description=FOOBAR
[Service]
Type=oneshot
ExecStart=/usr/bin/foobar.sh
RemainAfterExit=yes
[Install]
WantedBy=multi-user.target
# cat /usr/bin/foobar.sh
#! /bin/bash
. /etc/init.d/functions
echo_success
# LANG=C grep foobar.sh /var/log/messages
Aug 7 09:29:54 localhost foobar.sh[29607]: #033[60G[#033[1;32m OK #033[0;39m]#015#033[60G[#033[1;31mFAILED#033[0;39m]
This is a regression issue caused by the commit f88dbd98e992 where
consoletype command was dropped.
Under systemd service units, standard input is associated with
/dev/null by default:
# systemctl show -p StandardInput foobar.service
StandardInput=null
man 5 systemd.exec:
StandardInput=
Controls where file descriptor 0 (STDIN) of the executed
processes is connected to. Takes one of null, tty,
tty-force, tty-fail, data, file:path, socket or fd:name.
If null is selected, standard input will be connected to
/dev/null, i.e. all read attempts by the process will
result in immediate EOF.
consoletype command interprets /dev/null as serial:
# consoletype < /dev/null
warning: consoletype is now deprecated, and will be removed in the near future!
warning: use tty (1) instead! More info: 'man 1 tty'
serial
The commit f88dbd98e992 overlooks this situation.
To fix this issue, check also "not a tty" to cover the case where
standard input is /dev/null.
Note that LANG=C is needed because the message "not a tty" varies
depending on locales.
|
| |
|
|
| |
Resolve: BZ #1691552
|
| |
|
|
|
|
|
|
|
| |
In ifup/ifdown scripts, move deprecation waring info after
source_config, so users can config DEPRECATION_WARNING_ISSUED in
ifcfg-** file to decide whether show deprecation waring info when
calling ifup/ifdown.
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
|
| | |
|
| | |
|
| |
|
|
|
| |
Now with systemd, this does not work pretty well and users really should
use systemd unit-files for this.
|
| |
|
|
| |
Resolve problem when wasn't installed chkconfig package. BZ #1716386
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add option for time saving waiting
Replace date with simple timeout
Rename configurable variable and add documentation
Print log only when didn't reach IP address in time
Add support for IPv6 address
Remove unneeded output
Change -ge to -gt, because equal value is ok
Clarify DNS name option in documentation
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
IPv6 dhclient daemon only tries one time to obtain a IPv6 address
from a DHCPv6 server regardless of the setting of PERSISTENT_DHCLIENT.
PERSISTENT_DHCLIENT option is only used for IPv4 dhclient daemon.
With the popularization of IPv6 protocol, some users prefer setting
IPv6 like IPv4. I think there are two solutions as follows,
1. adopt PERSISTENT_DHCLIENT option to both IPv4 and IPv6.
2. create a new option, such as PERSISTENT_DHCLIENT_IPV6 option,
just for IPv6.
The first solution does not introduce addition options, and the IPv4
and IPv6 settings are consistent. So I perfer choosing the first
solution.
Fixes: bf00a0048 ("Replace /var/run with /run everywhere")
Signed-off-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
|
| |
|
|
|
|
|
|
|
|
| |
This was causing the /etc/resolv.conf file to be always updated when
RESOLV_MODS was not set...
Before the commit 5d6156454bf8f6dab4a5fdd7e1bf6 we were not updating
the /etc/resolv.conf file if the RESOLV_MODS was empty.
See https://bugzilla.redhat.com/show_bug.cgi?id=1610411 for more info.
|
| |
|
|
|
| |
GitHub displays the tag message when we create new release there,
so lets have message with more sense...
|
| |
|
|
|
|
|
|
| |
GitHub will create a new commit with different hash (merge) when
applying the pull-request, and so the tag ended up pointing to wrong
commit hash...
For now we have to tag the release commits manually.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Because the return value of 'convert2sec()' function can sometimes be
decimal, the follow up 'expr' call can fail, since 'expr' does not
support floating point calculations. This can sometimes lead to error:
"""
expr: non-integer argument
/etc/sysconfig/network-scripts/ifup-eth: line 91: [: 0: unary operator
expected
"""
To solve this bug, we switch to 'bc' utility, which supports floating
point computations. We also have to change the comparison condition of
$LINKDELAY and $forward_delay to use 'bc' as well.
|
| |
|
|
|
| |
In case of 'network' service these warnings are displayed only once,
to not spam unnecessarily user's journalctl if they have many NICs.
|
| |
|
|
|
|
|
| |
This partially reverts the commit 15eaf7ece, because systemd's
generated unit files might have incorrect dependencies at some
circumstances, thus causing diskless nfs-clients to hang while
unmounting root filesystem after '$ service network stop'.
|
| | |
|
| |
|
|
|
|
| |
This will allow services to install drop-in files into these folders
for the 'readonly-root' service, but that service won't have to be
installed by default as a result...
|
| | |
|
| | |
|
| |
|
|
|
| |
When converting from brctl to ip-link, the call to exit in case bridge
adding failed was dropped by accident.
|
| |
|
|
|
| |
Since ip-link has full support for Linux bridges (and slave ports), use
that instead of the deprecated brctl from bridge-utils.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
This makes sure all the subpackages stay installed during the upgrade
for Fedora users. After that users will be able to manually remove the
packages if needed...
Fresh installations will have only initscripts installed.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
To build in container-based environment because of speed.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Resolves: #75
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
For currently unknown reason the dbus-send calls will fail to set the
firewall zone for the given interface if we omit the --print-reply
option...
This looks like some kind of race-condition in dbus-send, since the
--print-reply makes the call synchronous and slower.
Hopefully this is only a temporary workaround until DBus is fixed.
Resolves: #1586284
|
| |
|
|
|
|
|
| |
This was causing the DBus call to fail, and we didn't catch it before
since we were forwarding everything into /dev/null...
Related: RHBZ#1586284
|
| | |
|
| | |
|
| | |
|
