Date: Tue, 21 Mar 2017 17:20:17 -0400 (EDT) From: "Igor R." <freebsd@str.komkon.org> To: Ernie Luzar <luzar722@gmail.com> Cc: freebsd-questions@freebsd.org Subject: Re: command line history broken in 11.0 Message-ID: <alpine.BSF.2.20.1703211656530.12389@tissa.komkon.org> In-Reply-To: <58D1925B.8050500@gmail.com> References: <alpine.BSF.2.20.1703211310410.12389@tissa.komkon.org> <58D1925B.8050500@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Ernie, You assertion about -o option sounds reasonable. I didn't think through if it is consistent with the changes made between the versions, and I didn't check the earlier history of changes (which would make your aliases+history working in 10.x) of the code. But the changes between 10.3 and 11.0 were indeed related to signal vs fork, which would be in the right ballpark with what the option -o does, according to the man pages. By the way, - I am still curious if sending halt or reboot into background would also do what you wanted. If you have a box you can experiment with, - would you mind doing that? As for deprecating halt(8) and reboot(8), - I think that is unreasonable. It is not about the time, but about how things are done. So, shutdown will run those (they are hardlinks) if -o specified, and send a signal to init(8) otherwise. You are thinking only about the situations when everything is working fine. I can see that in some cases init(8) might stop responding to signals, in which case halt/reboot *might* still do the job. Besides, as described in init(8): " When shutting down the machine, init will try to run the /etc/rc.shutdown script. This script can be used to cleanly terminate specific programs such as innd (the InterNetNews server). If this script does not terminate within 120 seconds, init will terminate it." Sometimes you might need to halt/reboot the system without running all of those shutdown scripts (while still flushing the file system). Besides, there is no downside of having them around (except for non-educated admins, but then the problem is at a different level altogether). Igor On Tue, 21 Mar 2017, Ernie Luzar wrote: > Igor R. wrote: >> >> It is not something that I've tried to do, but here is an idea: >> "reboot & ; exit". >> The rational is that by doing this you would detach the standard output >> of the "reboot" process, and that will allow the exit to complete. >> >> If that doesn't help, you can think about redirecting the stdout >> (and as the next step, - stderr) in that command. >> >> HTH! >> >> PS. I just looked at the source code, and I suspect that the difference in >> the behavior in 10.3 and 11.0 (or 11.0.1) might be related to these >> changes: >> Just before 10.3 release, this change was introduced to /bin/csh: >> https://svnweb.freebsd.org/base?view=revision&revision=296976 >> and then quickly reversed. by Revision 297204. >> https://svnweb.freebsd.org/base?view=revision&revision=297204 >> You can see that revision 296976 had weird side effects: >> "causes strange effects like for example SIGTERM not >> being delivered to rc(8) scripts on shutdown albeit these use sh(1), >> if csh(1) or tcsh(1) are used as login shell of root." >> >> >> At the same time, soon after 10.3-release was split from HEAD, that change >> was reversed in HEAD as well, but in addition to that (Revision 297672) a >> different call (fork) was introduced: >> https://svnweb.freebsd.org/base?view=revision&revision=297673 >> https://svnweb.freebsd.org/base/head/bin/csh/config_p.h?r1=297673&r2=297672&pathrev=297673 >> >> Maybe that change also had some side effects (like what you see), but just >> less severe or less noticeable, - so it "worked" for 11.0*-release. >> >> Assuming my guess is correct, I'd recommend you filing a bug report (pr), >> so that it is documented - and that can bring it to the attention of the >> core team. >> https://bugs.freebsd.org/bugzilla/enter_bug.cgi >> >> >> >> On Mon, 20 Mar 2017 14:05-0400, Ernie Luzar wrote: >> >>> On 10.3 I had the current session saving the command line history when >>> issuing >>> the shutdown, halt, and reboot command by using these alias commands that >>> I >>> added to the .cshrc file of my logged in user account. >>> >>> alias sd "exit && shutdown now" >>> alias sdp "exit && shutdown -p now" >>> alias rboot "exit && reboot" >>> alias stop "exit && halt" >>> >>> Now after doing a clean install of 11.0 and using the same .cshrc file the >>> rboot and stop alias commands no longer save the current history. They act >>> like the exit command is not getting executed. The sd and sdp alias >>> commands >>> are working as expected. >>> >>> When existing from a session terminal by issuing the exit command does >>> still >>> save the current history. >>> >>> Is there an alternate method I can use? > > > Thank you Igor. > > Your reply hit the mail on the head. You have proved there is a reason for > the different between 10.3 and 11.0. I now know its not me screwing something > up. > > I've been a Freebsd user since release 3.0. Over time I noticed that the > shell command line history was NEVER being updated so I would see the history > from the previous session merged into the old .history file the next time I > started the host. It was not existing the shell and logging on the next day, > instead I would issue halt or shutdown -p now command. > > During the 10.x series some time I posted a question to the questions list > about this behavior. To make a long story short, the solution at that time > was to do it this way and it was working fine. > > alias rboot "exit && reboot" > alias stop "exit && halt" > > Then along comes 11.0 and it stopped working, I tried many different > combinations and nothing worked. It was Polytropon who gave me the pointer I > needed. Even though I had reread the shutdown man page a few times, I never > really understood what the meaning of the -o flag was really talking about. I > recoded my alias commands like this > > alias sd "shutdown -o now" > alias sdp "shutdown -po now" > alias rboot "shoutdown -ro now" > alias stop "shutdown -ho now" > > Low and be hold, Now the shell is given a chance to close down normally which > allows the shell .history file to be updated. My world is happy again. This > also explains why sometimes at shutdown my database gets corrupted > > To my way of thinking this -o should be the default behavior instead of how > it is now. This default behavior is now in violation of the original > intention for using the shutdown command in the first place, which is to > allow running tasks to close down cleanly saving databases from becoming > corrupted. > > It would be best if halt and reboot commands were deprecated and removed from > the base system leaving the shutdown command the only way to bring down the > system, aside from the power button. The shutdown -o mode should become the > default mode. Then "shutdown -r now" becomes the replacement for the "reboot" > command and "shutdown -h now" becomes the replacement for "halt" command. > Their is never a technical need where it's necessary to just terminate > everything thats running. This concept of terminating all running tasks is a > legacy leftover from the first BSD version that now needs to be corrected. > With todays powerful computers there is no noticeable lag in allowing tasks > to shutdown cleanly. > > To address people who may have reboot or halt inbedded in some scripts, the > reboot command could be an alias for "shutdown -r now" and halt an alias for > "shutdown -h now". > > I think a PR is needed here to get this problem documented. It surely effects > every user of FreeBSD. > > Now if I have missed something or mis-understood something, I sure would like > to hear about it before I submit a PR on this. Even an acknowledgment that I > got it correct would be good to hear. > > Thanks to everyone who replied. > > >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.20.1703211656530.12389>