Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
Hi, Upstream of the procps-ng adopted removing -v from pkill, while keeping it for pgrep. The change allows one to reach the inversion logic with long option e.g. 'pkill --invert' but that is not documented (while it perhaps should be). https://gitorious.org/procps/procps/commit/1af18c260a87dc38f0e33bfeb6de6163f91be4ad -- Sami Kerola http://www.iki.fi/kerolasa/ -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
Hi Jason, On Thu, Feb 04, 2010 at 04:01:33AM -0500, Jason A. Spiro wrote: Thanks for writing NetBSD pgrep/pkill. When you have some spare time to carefully read ten messages and think, you please read through http://bugs.debian.org/558044 and suggest what you think would be a good algorithm for making -v safer? I would chalk this down to experience. In my youth I typed killall on an ICL DRS6000 running SVR4 on the assumption that it would kill one process. In fact it brought down the entire system, as designed, leaving me very red faced. So personally, having made the mistake and learned to be very careful about killing things I don't see it as a big issue, although perhaps there is a good way to improve it and make it safer. Unfortunately, I don't have any good ideas in this area. If you'd like to solicit ideas from the greater NetBSD userbase, tech-userle...@netbsd.org would be a good place to ask. They can be a noisy and opinionated bunch so liberal use of your mail client's Delete facility would be advised, if you chose to go down this route. :-) Note that the suggestion of disallowing -v with a pattern could be problematic because it would prevent rare and strange, albeit meaningful constructs such as: pkill -v (init|syslogd|sh|randomprogram) Thank you for asking my opinion. Cheers, Andrew -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
Hi Andrew, Thanks for writing NetBSD pgrep/pkill. When you have some spare time to carefully read ten messages and think, you please read through http://bugs.debian.org/558044 and suggest what you think would be a good algorithm for making -v safer? Thanks in advance, -Jason -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: changing pkill -v
as the original author of Linux' pgrep, I support the idea of making the use of -v an error unless a search criterium other than pattern is specified. (-f is not a search criterium in itself.) there is a sort of precedent for this, since -o and -n has always been disallowed with -v since the behaviour would be useless. wrt. adding a verbose mode for pkill I think it would be best done by adding -l, even if other kill programs use that option to list signal names. but that's not pertinent to this change request :-) -- best wishes, Kjetil T. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044:
On Thu, Nov 26, 2009 at 6:41 AM, Craig Small csm...@debian.org wrote: I disagree. pgrep/pkill (they're the same binary) are like grep and use the same flags. I think that is important. I think it is more important to protect users. But you are the maintainer and so the decision is up to you. Now I will continue to respond to the other arguments that you said in your last mail. Changing the flags now will cause more problems than it will solve because now you will have different flags for new Debian servers, old ones and the rest of the world. Maybe you could gradually transition it in, by making both options work for a year or two and printing a warning message, then disabling -v. If you are not careful you can do bad things easily. pkill -v is in the same vein as fuser -km / or the old rm -rf / mydir/myfile. Yes. But I think it would be good if you disable this particular bad thing, just like GNU disabled the old rm -rf / mydir/myfile. Principle of least astonishment is all good, but inconsistent flags across different Linuxs is even worse. Which Linux does not use Debian procps? For example, it looks like Fedora uses Debian procps.[1] While I do understand what you are trying to do here I don't agree. Given that, I'm closing this bug. What would you think if we posted on the debian-devel list and asked for more advice on other possible options to prevent accidental killing-all-processes with pkill -v? ^ [1]. http://cvs.fedoraproject.org/viewvc/rpms/procps/devel/FAQ?view=markup -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: Fwd: Re: Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
-- Forwarded message -- From: Craig Small csm...@enc.com.au Date: Wed, Dec 2, 2009 at 3:06 AM Subject: Re: Bug#558044: To: Jason A. Spiro jasonspi...@gmail.com On Wed, Dec 02, 2009 at 02:59:38AM -0500, Jason A. Spiro wrote: Maybe you could gradually transition it in, by making both options work for a year or two and printing a warning message, then disabling -v. That's been done before, there were lots of complaints. Which Linux does not use Debian procps? For example, it looks like Fedora uses Debian procps.[1] They copied the FAQ from upstream, I saw nothing about it being a Debian specific patched version. What would you think if we posted on the debian-devel list and asked for more advice on other possible options to prevent accidental killing-all-processes with pkill -v? If the bulk of debian-devel said yes this is a good idea I'd regretfully change it. Now the rm -rf / blah protection is quite neat and perhaps something like it could be done. I'm not sure what the criteria would be. - Craig -- Craig Small GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5 http://www.enc.com.au/ csmall at : enc.com.au http://www.debian.org/ Debian GNU/Linux, software should be Free -- Jason Spiro: software/web developer, packager, trainer, IT consultant. I support Linux, UNIX, Windows, and more. Contact me to discuss your needs. +1 (416) 992-3445 / www.jspiro.com -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
On Wed, Dec 2, 2009 at 3:06 AM, Craig Small csm...@enc.com.au wrote: That's been done before, there were lots of complaints. When? What were the complaints? :) They copied the FAQ from upstream, I saw nothing about it being a Debian specific patched version. But Craig, aren't you the upstream maintainer of procps? If the bulk of debian-devel said yes this is a good idea I'd regretfully change it. Thank you for offering to do so. Now the rm -rf / blah protection is quite neat and perhaps something like it could be done. I'm not sure what the criteria would be. Good idea. I have researched your idea. From the first forty or so Google search results for [ pkill -v ], it looks like nobody in those search results uses pkill -v some_process_name except by accident. They only use -v purposefully when they use it with -u, as in pkill -v -u root or pkill -v -u 0,1 or pkill -v -u root,daemon,nobody,gdm. Maybe if you want to use -v without -u, you should need to pass the long option --invert-match instead of the short option -v. What do you think? -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: Fwd: Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
-- Forwarded message -- From: Craig Small csm...@enc.com.au Date: Wed, Dec 2, 2009 at 5:14 AM Subject: Re: Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose To: Jason A. Spiro jasonspi...@gmail.com On Wed, Dec 02, 2009 at 04:09:30AM -0500, Jason A. Spiro wrote: When? What were the complaints? :) It was quite a few a years ago and it was basically we want it the other way But Craig, aren't you the upstream maintainer of procps? No. Albert is. I'm the Debian procps maintainer. I can only generate Debian-specific diffs, so thats why changing the flags is a bad idea. I'm also the psmisc (eg killall) maintainer. Good idea. I have researched your idea. From the first forty or so Google search results for [ pkill -v ], it looks like nobody in those search results uses pkill -v some_process_name except by accident. They only use -v purposefully when they use it with -u, as in pkill -v -u root or pkill -v -u 0,1 or pkill -v -u root,daemon,nobody,gdm. Maybe if you want to use -v without -u, you should need to pass the long option --invert-match instead of the short option -v. What do you think? Something like that would be better. Not changing the flag but disabling it for somewhat obvious errors; and then having another flag, a long option, that means yes really do it. - Craig -- Craig Small GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5 http://www.enc.com.au/ csmall at : enc.com.au http://www.debian.org/ Debian GNU/Linux, software should be Free -- Jason Spiro: software/web developer, packager, trainer, IT consultant. I support Linux, UNIX, Windows, and more. Contact me to discuss your needs. +1 (416) 992-3445 / www.jspiro.com -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
cc added: Albert Cahalan acahalan %at% gmail.com. Albert, welcome to the thread, and thanks for maintaining procps. On Wed, Dec 2, 2009 at 5:14 AM, Craig Small csm...@enc.com.au wrote: On Wed, Dec 02, 2009 at 04:09:30AM -0500, Jason A. Spiro wrote: When? What were the complaints? :) It was quite a few a years ago and it was basically we want it the other way What functionality were you gradually transitioning in? But Craig, aren't you the upstream maintainer of procps? No. Albert is. I'm the Debian procps maintainer. I can only generate Debian-specific diffs, so thats why changing the flags is a bad idea. I'm also the psmisc (eg killall) maintainer. Ah. Craig, when I saw You may use the Debian bug tracking system at http://procps.sourceforge.net/faq.html , and you replied to my issue mail, I thought you were the upstream. :) Is it allowed in the current version of debbugs for you to configure things so debbugs will forward all new procps bugs not just to you, but also to Albert? Good idea. I have researched your idea. From the first forty or so Google search results for [ pkill -v ], it looks like nobody in those search results uses pkill -v some_process_name except by accident. They only use -v purposefully when they use it with -u, as in pkill -v -u root or pkill -v -u 0,1 or pkill -v -u root,daemon,nobody,gdm. Maybe if you want to use -v without -u, you should need to pass the long option --invert-match instead of the short option -v. What do you think? Something like that would be better. Not changing the flag but disabling it for somewhat obvious errors; and then having another flag, a long option, that means yes really do it. Albert, I wrote an explanation -- viewable at the top of http://bugs.debian.org/558044 -- on why pkill -v is so dangerous. Given my explanation, what do you think of my idea? Albert, Craig, please make sure to always CC me and the bugtracker when you reply. Kind regards, -Jason -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: reopening
reopen 558044 = thankyou The current way of using pgrep -v can often lead to a large number of processes killed and often this is not the intention of the user. A way of limiting the damage needs to be added to pkill without changing the flag. Some ideas are to disable -v if no process (before the negation) is found or to not by default be able to use -v if no other flags are used. In other words pkill -v myproc will not work but kill -v myproc -u root will and pkill --some-flag myproc will. -- Craig Small GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5 http://www.enc.com.au/ csmall at : enc.com.au http://www.debian.org/ Debian GNU/Linux, software should be Free -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044:
I see from http://www.cyberciti.biz/tips/my-10-unix-command-line-mistakes.html#comment-150923 that pkill's options are derived from pgrep's which are derived from grep's. But still, the benefits of preserving grep-like -v behavior don't outweigh the costs. The benefits of preserving Solaris compatibility for -v aren't worth the costs either. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#558044: pkill, pgrep: don't use -v for negation; it's dangerous some think -v means verbose
Package: procps Version: 1:3.2.8-2 Severity: wishlist X-Debbugs-CC: mo...@mozai.com;volker.maib...@eberspaecher.com Thank you for maintaining procps. A suggestion: killall -v Foo means kill process Foo, and print verbose output. Therefore many users think that pkill -v Foo means the same thing.[1][2][3][4][5][6][7] It doesn't. It kills all processes except Foo. This tends to cause harm. Please follow the Rule of Least Surprise. Don't allow -v to mean the more destructive of its two possible meanings. Instead, please rename the -v option to some other letter. Not -i for invert: killall uses -i to mean interactive. And preferably not -L: skill uses that to List all possible signals. Use some other letter. Preferably also make pkill -v print a special error message telling you the new way to negate your criteria. Please change pgrep too, so that its option names will match pkill's. ^ [1]. It's ... good to know that pkill -v doesn't give you verbose output. --http://www.reddit.com/r/AskReddit/comments/7vkpi/whats_the_worst_mistake_youve_ever_made_at_your/c07j2yv ^ [2]. Just assuming that a [-v] option stands for --verbose. Yep, most of the time. But not on a [pkill] command. --http://www.cyberciti.biz/tips/my-10-unix-command-line-mistakes.html#comment-149235 ^ [3]. I used that once to kill all evolution processes on my machine: 'pkill -v evolution' ... believe me, next time I will read the man-pages before I try an unknown option :-). --http://www.mail-archive.com/ads...@vm.marist.edu/msg62476.html ^ [4]. Note to self: the '-v' in 'pkill -v $pid' does not mean --verbose. --http://www.df7cb.de/blog/2007/pkill.html ^ [5]. It turns out 'pkill -v x' is NOT 'verbose'; rather, it's 'kill all processes except x'. Kind of a bummer! --http://twitter.com/kellyterryjones/status/4112024256 ^ [6]. `pkill -v rsync` does NOT mean 'kill all rsync processes and be verbose about it.' shamehangs head/shame --http://twitter.com/mozai/status/5240550425 ^ [7]. Many more examples are available if you do a Google search for [ pkill -v ]. -- Jason Spiro: software/web developer, packager, trainer, IT consultant. I support Linux, UNIX, Windows, and more. Contact me to discuss your needs. +1 (416) 992-3445 / www.jspiro.com -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
