I always have /proc/sys/kernel/panic set up to 0. Looking at description of this option in kernel.org we can see:
panic:
The value in this file represents the number of seconds the kernel
waits before rebooting on a panic. When you use the software watchdog,
the recommended setting i...
If sources.debian.org/stats is accurate, the number of source packages in Debian is approaching 30k. When I first started using Debian, around potato, it was slightly under 3k.
But the rate of growth seems to be slowing. Though I suppose that's not surprising.
@StephenKitt the mod team's discussing whether we should delete this answer as too dangerous and likely to break your system. Since you're our resident Debian expert, what do you think?
Editorial note:
Please read the comments to this answer.
Debian does not have Python 3.6 in its repositories, but testing has it.
$ sudo nano /etc/apt/sources.list
# add
deb http://ftp.de.debian.org/debian testing main
$ echo 'APT::Default-Release "stable";' | sudo tee -a /etc/apt/apt.conf.d...
that’s really the general approach for many, many “I want newer foo” questions: the packages are there for the system, if you’re developing, you should use your development environment’s package management and its virtual environment system
and then deploy using containers but I may have been infected by this whole cloud malarky
Has Debian changed to be more dangerous with regard to pulling packages from 'testing' lately? I don't remember it going horribly wrong before. It's not that you could usefully pull just a single package though, as they'll pull newer versions of libraries etc.
I thought the idea was to allow people to post their answers and others to vote on them... (not that it always ends with the right solution, but still)
As of 9/15/2018 it is still risky: I tried this on a Raspberry Pi with Debian stretch and ran into segmentation faults after the Python 3.6 upgrade from testing pulled in an updated libc6 and libc6-dev. — drottSep 15 '18 at 7:28
@ilkkachu perhaps a better solution here would be to turn the A into a community wiki, and rewrite it so it’s correct; the A’s author is a one-shot wonder with no participation before or since
Sure, there are better answers: starting from seeing if the newer Debian packages would compile against the old version. Or compiling it from source. Or upgrading the whole thing to testing or unstable, so that you at least have a somewhat consistent set of packages (not that running them was actually supported either, IIRC). Or installing some other distro that packages what they wanted. But deleting that just because people didn't read the warning on top of the answer?? :D
Just to add my two cents here, and I'm biased from being in the LQP queue regularly, the A attempts to answer the question. IMHO, the edits and comments frame it. Votes should account for the rest. (Yes, the mod team has opinions) :)
BUT, if there is a reasonable chance of its breaking a user's system in ways most users can't recover from, then I would argue we don't need it around. Especially when there are other, better and safer methods posted.
@terdon For the sake of argument, let's say "above average", though I'm Finnish and loathe to write even that. But where does the line go? Should we allow for different solutions to be presented, or draw a line somewhere about solutions that require "too much experience"?
Yes, the fact that it's the accepted answer, and on top, and doesn't provide other solutions is... not a very good thing.
@ilkkachu I would say that requiring experience to manage to implement the solution is fine. But something that is trivial to do BUT will break your system in ways that will require experience to fix it is a different thing.
But let's get back to this. Our objective here is to build a library of good solutions to *nix problems. With that in mind, and considering that we have several answers giving better approaches, what are the arguments in favor of keeping an answer that is at best sub-optimal and at worst actively harmful?
@JeffSchaller There are probably suboptimal accepted answers around the site, too. Partly because people tend to accept the first one, and better ones take longer to write.
So, I see three issues here. 1) Is this answer really dangerous enough to justify deletion? 2) Do we really want mods to go around deleting answers they consider dangerous?
I think the answer to 2 is a clear "no" unless the issue is blindingly obvious (e.g. posting rm -rf --no-preserve-root as an answer to a random question).
More seriously, what do you mean other options? Non-accepted answers can be downvoted until they're <0 and then +20k users can vote to delete them. Accepted answers, however, can only be dealt with by mods.
We had another situation like this a while back where the OP had accepted a very wrong answer, which couldn't be deleted because it was accepted and was just accumulating downvotes. The answerer ended up flagging and asking us to delete since they couldn't do it themselves.
@terdon anything. I might not know them all. Editing being the obvious one. Adding banners with warnings and links to better answers. Hiding the bad answer in a spoiler block. I don't know :(
@StephenKitt Yes. But that, again, is something I'm not comfortable doing. I might if it were on a subject I was expert enough in to be sure my answer would be 100% correct, but not on this one.
@ilkkachu Ah right. Yes, adding a warning is done often enough.
I accept that the answer there shouldn't be on the top, I don't like the idea of massive edits, but deleting really feels like something my gut opposes. :/
@terdon It's possible (and not even difficult), to install more recent versions of Python alongside the official versoins. If it is done carefully, it's completely safe. I think I might have done this once or twice. The main thing you have to watch out for is to avoid making that more recent version the default. If you do that, then of course bad things will happen. But I think this is dealt with by suitable tweaking of the build files. And of course, it will need to be rebuilt on the system.
@JeffSchaller which is what I did ;-). But I thought answers were supposed to stand alone, i.e. still mean something even if all the other answers are deleted.
@terdon Yes I did just wonder about that, too :) Partly because it's not the mods picking and choosing. Of course, in this case it would be a bit of a cop-out, since we'd use another excuse to delete the whole lot basically just to avoid the messy thing.
Python 3.6 (though rc1) just arrived in Debian experimental. I was able to successfully build this on Debian stable (jessie). See the final steps of the build pasted below.
However, that doesn't mean it will install successfully and not cause problems. A reasonable next step would be to test it ...
@StephenKitt sure; it says "(misdirected stuff), oops I misunderstood, (more stuff)". If they self-deleted it fine. If there was pressure on me to delete it, I'd consider it (and whether I felt comfortable enough deciding it wasn't directed at the Q at all any more)
@FaheemMitha oh, you mean installing packages? yes, that works, but it’s rarely supported — stable releases are only supposed to have a single version of each major release of Python
so parallel installs work because that’s how we drive transitions
@FaheemMitha the infrastructure is still there, yes, and it’s still used, but I don’t think there were ever that many versions in parallel in a stable release
currently testing has 3.7 and 3.8
but stable only has 3.7
back to the question at hand: @terdon, I think the first comment should be deleted anyway, it’s obsolete and misleading
(thanks to cas’ edit)
the second comment too, since it mentions experimental
@StephenKitt There were, but it was a long time ago.
I remember because at some point I went in there and had to manually purge all the old python versions and libraries that nothing was using anymore. Similarly for gcc and accompanying libraries.
Curious, when you gurus write programs, do you always write its dependencies? So for example, when you upgrade your OS and libraries, everything is broken, and you have to either rewrite your programs or look for the particular versions of dependencies to install? Or do you keep faith in the backward compatibility of the dependencies?
And ideally programs are not super sensitive to dependency versions. Though sometimes it is necessary (or just preferable) to write for the bleeding edge, like when an ecosystem is evolving rapidly.
Biber is an example of a program that annoyingly insists on using the most recent version of Perl. 5, I think. I'm not a Perl user, but apparently it has something to do with evolving multilingual support.
But it's a pain when one is trying to backport.
Currently perl (>= 5.26).
I thought Perl 5 was a maintenance branch. Isn't there a Perl 6 now?
I read a little about multivariate statistics, nonparametric statistics, time series, bayesian statistics, but didn't get to understand well before being forced to stop
I am now wondering if Rust and Julia are better languages than C++ and Python or perl 6. That makes me interested in knowing the two sometime (not sure when)
Not vehemently or anything, we had a very nice, constructive discussion. As far as I'm concerned, since the reaction wasn't a resounding "yes, absolutely, delete it", I won't.
@ilkkachu It's not dangerous. But you need to be careful, and need to know what you are doing. And you also often need to rebuild locally. But not always. I backport stuff all the time. It's my standard way of getting newer releases.
You can't just install binaries that are meant for a newer version of a system, and expect everything to be fine. Not on Unix-like systems, at least.
The current shared library model has its limitations.
@ilkkachu Backporting is often possible, unless there are too many dependencies against newer versions. Then you end up going into an unpleasant recursive loop.
But bear in mind that the dependencies that Debian packages declare are often not sharp. So one can often get away with older versions.
@terdon, re that debian python thing, I see Stephen added a bigger warning, but was there a way for mods to add some special banner to posts? I have a vague idea of seeing such in some stack, but I'm not sure
The question has a downvote for some reason. Weird.
I made the answer CW, which I kind of regret. Because someone made an unfortunate edit. I probably should fix it.
I don't know if anyone here has followed the TV version of "The Magicians", which recently ended after 5 seasons.
I just finished watching the first season. They really didn't follow the books closely at all.
Normally that would not be a good thing, but in this case, the script writers are clearly very invested in the material, and thought they did a good job with their modifications.
The casting is also pretty good.
They completely threw out the dramatic climax of Book 1, though. Maybe they'll use it later.
Huh, just noticed that they only have the first two seasons on Amazon Prime Video.
Maybe they removed it. I thought they had them all.
I always have /proc/sys/kernel/panic set up to 0. Looking at description of this option in kernel.org we can see:
panic:
The value in this file represents the number of seconds the kernel
waits before rebooting on a panic. When you use the software watchdog,
the recommended setting i...
Editorial note:
Please read the comments to this answer.
Debian does not have Python 3.6 in its repositories, but testing has it.
$ sudo nano /etc/apt/sources.list
# add
deb http://ftp.de.debian.org/debian testing main
$ echo 'APT::Default-Release "stable";' | sudo tee -a /etc/apt/apt.conf.d...
