|
other people posted:supervisord s6 or bust.
|
# ? Nov 18, 2018 12:13 |
|
|
# ? May 2, 2024 12:52 |
|
Hi goons - I'm working on a project using Ansible. Once I'm done setting up my server, I want to take some "screenshots" of the state of the server. These are just commands like top/df/free, but I'd like to capture them including the command and prompt to use in documentation later. Any tips on how to achieve this? I'm okay just dumping text and then converting it to a screenshot.
|
# ? Nov 19, 2018 14:11 |
|
Gyshall posted:Hi goons - I'm working on a project using Ansible. You could have a look at scrot.
|
# ? Nov 19, 2018 16:36 |
|
Or asciinema. That's what I use to playback a playbook.
|
# ? Nov 19, 2018 22:20 |
|
Sheep posted:Trying to do a CentOS install via kickstart on some wonky machines with NVIDIA cards; the text installer (it's all automated) seems to be dying at some point in the install on these machines but I can't tell because the screen is garbled. First time I've seen something like this in text mode. Maybe you could try getting it to use a serial console - perhaps it won't make the install work, but you'll at least be able to see where it went wrong? my bitter bi rival posted:
For the record, that looks like Perl, but since you already have a working solution there's probably no point in suffering through turning it into a slightly smaller one in Perl other people posted:RHEL 8 Beta is out: This is what I came here and caught up on the last few pages of the thread for. It's true, there's no KDE I really don't think I want to get to know how GNOME 3 works. Are there any other options that don't involve moving to another distribution? Using KDE on CentOS in the past meant I didn't always have all the config tools, but I didn't mind running some GNOME/GTK-based config tools under KDE, and I wouldn't mind having to run some of them under some other window manager or desktop environment on CentOS 8.x. I suppose by the time CentOS 7.x is no longer supported, EPEL might include KDE for CentOS 8.x if I'm lucky. Or maybe IBM will prevent CentOS 8.x somehow.
|
# ? Dec 9, 2018 07:33 |
|
Buttcoin purse posted:Maybe you could try getting it to use a serial console - perhaps it won't make the install work, but you'll at least be able to see where it went wrong? Don't use RHEL/CentOS on the desktop, use Fedora.
|
# ? Dec 9, 2018 13:14 |
|
What is the execute bit actually used for in linux? I learned in school that it's required to have the system run a binary, but what is the practical use case for it? I was thinking today that a user who can write to any file with the +x bit can just copy binary contents and then run the code they want as if it originally had the +x bit: [root]# chmod 444 ls [root]# logout [idiot]$ ./ls -bash: ./ls: Permission denied [idiot]$ touch new_ls [idiot]$ chmod 744 new_ls [idiot]$ cat ls > new_ls [idiot]$ ./new_ls /dev | head -1 autofs Why isn't the ability to execute system binaries then a property of the user instead of the file? Is there anything different about executing this binary owned by idiot instead of one owned by root?
|
# ? Dec 9, 2018 21:25 |
|
Salt Fish posted:What is the execute bit actually used for in linux? I learned in school that it's required to have the system run a binary, but what is the practical use case for it? I was thinking today that a user who can write to any file with the +x bit can just copy binary contents and then run the code they want as if it originally had the +x bit: Don't think too hard about it. remember that the basic unix permissions systems predates concepts like "effective security". Add selinux and it starts getting better. In more practical terms, don't give anyone read access to anything the shouldn't execute, but if they do pull a trick like that, the executable is still only running with their permissions. To break out of that would need them to be able to set suid.
|
# ? Dec 9, 2018 22:12 |
|
Somewhat of a tangent: I had a Google SRE interview question like "There's a huge binary on a system that you need to run, but it's owned by someone else and missing the +x flag. It's too huge to copy to a new file. How do you execute it?" That led to a fascinating rabbit hole of understanding the way unix permissions and program execution works. Like, could you make a soft/hard-link to the file and the link has the execute bit? Could you write a wrapper program that just does a syscall to "exec /path/to/file"? What about loading it in a debugger? etc.
|
# ? Dec 9, 2018 23:50 |
|
You can run any ELF executable with ld-linux.so. In Debian it's named something like /lib64/ld-linux-x86-64.so.2 .
|
# ? Dec 10, 2018 00:20 |
|
waffle iron posted:You can run any ELF executable with ld-linux.so. In Debian it's named something like /lib64/ld-linux-x86-64.so.2 . This
|
# ? Dec 10, 2018 01:48 |
|
RFC2324 posted:Don't use RHEL/CentOS on the desktop, use Fedora. But I don't want to be a beta tester and I don't want everything to be broken once a year, I want stable software and to only have a major upgrade once a decade or so. Maybe I should learn Debian and try out their LTS releases, I gather they're supported for 5 years? Perhaps to make up for the sin of not joining in the collective beta testing effort, I could run Fedora in a VM and do.. something with it. I already run some things under Fedora but I really want important stuff like the machine booting and my email and web browser running to just work. Salt Fish posted:What is the execute bit actually used for in linux? I learned in school that it's required to have the system run a binary, but what is the practical use case for it? I was thinking today that a user who can write to any file with the +x bit can just copy binary contents and then run the code they want as if it originally had the +x bit: One thing you can perhaps do is prevent the user from writing anywhere other than their home directory by not putting them into the necessary groups and not having any directories outside /home that are world-writable, and then have /home mounted with the "noexec" flag: code:
|
# ? Dec 10, 2018 03:14 |
|
Salt Fish posted:What is the execute bit actually used for in linux? I learned in school that it's required to have the system run a binary, but what is the practical use case for it? I was thinking today that a user who can write to any file with the +x bit can just copy binary contents and then run the code they want as if it originally had the +x bit:
|
# ? Dec 10, 2018 03:21 |
|
I am trying to figure out how to enable the ability to right click by tapping an holding my touchscreen on my laptop. It is a fairly fresh Manjaro install with xfce on an old lenovo flex. I found recommendations for touchegg, which seems to do nothing. And for mousetweaks which successfully activates holding left click to right click, but only for the trackpad. It totally ignores the touchscreen. Anybody know of other programs to try, or hidden options to get those programs to work.
|
# ? Dec 10, 2018 15:34 |
|
Oh, there's a new major release for FreeBSD. Going to find out how easy the upgrade process is with this thing.
|
# ? Dec 13, 2018 10:35 |
|
code:
users with root access and a cron job
|
# ? Dec 13, 2018 22:35 |
|
xzzy posted:
How long did df or mount take to query?
|
# ? Dec 14, 2018 03:14 |
|
Can someone teach me the computer I am trying to set up a python script which is all good, except I have no idea how to get from manually navigating to the .py file and using "python undervolt.py", to where it's a universal command that you can just type in anywhere (like stuff that gets installed with the package manager). When he sets it to start as a service in his examples, he just writes "ExecStart=undervolt -params" without the whole path or running python. How do I do that? I'm sure the whole path would work, I just want to learn how to do it properly and I don't even know how to describe it well enough to Google it.
|
# ? Dec 14, 2018 11:22 |
|
The first step is to have the interpreter line in the first line of the undervolt.py file. It makes it so that you can omit typing python before undervolt. You also need to mark the file as executable for that to work. The pip script/package manager copies the file into your PATH, so that you can omit the rest of the path. On my system that would be a user thing ~/anaconda/bin/undervolt ,for example. The .py gets omitted in the same step. The whole process is normally called installing, and I agree that it is annoying to google and to learn the details on how to do it correctly without blindly using automated tools. btw. after you have installed using pip, you can check where it went with "which undervolt". And then open the file to check the interpreter path. VictualSquid fucked around with this message at 11:36 on Dec 14, 2018 |
# ? Dec 14, 2018 11:33 |
|
Thanks, that makes things clearer. I did try the pip install and still wasn't working ('which undervolt' returned nothing), but it works now that I've rebooted... so I guess it indeed added the PATH, and just didn't restart whatever service that loads those?
|
# ? Dec 14, 2018 11:42 |
|
That is a bit strange. To change the PATH you need to re-login or reboot, but not to install something. The PATH variable is a list of directories where your normal executable files live. You can check it with echo $PATH. It shouldn't change unless you are making major changes to you system. If pip just needs to copy something there you shouldn't need to restart. It might be possible that pip had to change some settings, which only makes sense if you are running pip for the first time.
|
# ? Dec 14, 2018 13:20 |
|
I remember when I used to use tcsh, you had to run the command rehash for it to see new binaries in the PATH. Is that still a thing in any shells? I've been using bash for a long time now.
|
# ? Dec 14, 2018 13:56 |
|
bash will find new binaries in your path, but if you have something in /usr/local/bin (or another location that's higher priority) and usr/bin, and delete it from local it'll still try to run the now missing one. i think that's the only case i've found recently where it doesn't just work correctly.
|
# ? Dec 14, 2018 15:51 |
|
It’s probably best to do a hash -r, just to be safe, especially if you’re about to build something that might care which version of a command it’s running (for instance, I’ve had builds fail because BSD sed and GNU sed produced different outputs).
|
# ? Dec 14, 2018 18:06 |
|
Buttcoin purse posted:I remember when I used to use tcsh, you had to run the command rehash for it to see new binaries in the PATH. Is that still a thing in any shells? I've been using bash for a long time now. Zsh needs rehash for completion. But I doubt any shell needs it to just execute, as ‘discovery’ handled at a lower level.
|
# ? Dec 14, 2018 23:05 |
|
Truga posted:bash will find new binaries in your path, but if you have something in /usr/local/bin (or another location that's higher priority) and usr/bin, and delete it from local it'll still try to run the now missing one.
|
# ? Dec 14, 2018 23:24 |
|
Hey dudes; looking for some advice on what usually comes with Linux. I'm building a frontend framework with Rust, and am working on Windows. I set up the Windows Subsystem for Linux to test (Ubuntu without GUI inside Windows), and ran into roadblocks installing Rust packages. Specifically, errors until running sudo apt update, installing GCC, Build-essential, OpenSSl, libssl etc, all which can be done using apt. I got around these by installing the packages, deduced either from the errors or Googling, but this could be a frustrating experience for someone not used to Rust. Are these normally included in Linux distros like Ubuntu, and omitted from the version that comes with Windows? Also, what's the usual convention for calling python in Linux? Eg in Windows you'd use the `python` command, but in the Subsyst for Linux, the correct command is `python3`. Also running into an issue where you need to run `chmod +x build.sh` on a build script to authorize its execution; is this normal as well? Trying to make this process as painless as possible, and document it. Is it assumed that Linux users would know how to deal with these issues offhand? Should I include a script that runs apt install for these packages, and chmods the build script? Dominoes fucked around with this message at 22:23 on Dec 18, 2018 |
# ? Dec 18, 2018 22:14 |
|
Dominoes posted:Hey dudes; looking for some advice on what usually comes with Linux. I'm building a frontend framework with Rust, and am working on Windows. I set up the Windows Subsystem for Linux (Ubuntu without GUI inside Windows), and am running into some roadblocks installing Rust packages. Specifically, errors until installing GCC, Build-essential, OpenSSl, libssl etc. I was able to get around these by installing the right packages, either from the errors or Googling, but this would be a frustrating experience for someone not used to Rust. Are these normally included in Linux distros like Ubuntu, and ommitted from the version that comes with Windows? Also, what's the usual convention for calling Python in Linux? I'm running into issues of `python` vs `python3`. Eg in Windows you'd use the `python` command, but in the Subsyst for Linux, `python` runs and old version, and you have to run `python3`. On the other hand, if you install stuff through the package manager it will install the dependencies for you. Ubuntu uses python3 for python 3 for compatibility reasons. WSL is basically just normal Ubuntu so you can probably find answers to questions by googling for information about Ubuntu rather than wsl specifically.
|
# ? Dec 18, 2018 22:24 |
|
mystes posted:If you're installing stuff manually you have to make sure the dependencies are installed yourself; Ubuntu doesn't magically know you want to use rust any more than Cargo will magically know your program's dependencies unless you tell it. You can probably find instructions for what you need installed. Dominoes fucked around with this message at 22:32 on Dec 18, 2018 |
# ? Dec 18, 2018 22:28 |
|
Dominoes posted:Specifically, errors until running sudo apt update, installing GCC, Build-essential, OpenSSl, libssl etc, all which can be done using apt. I got around these by installing the packages, deduced either from the errors or Googling, but this could be a frustrating experience for someone not used to Rust. Are these normally included in Linux distros like Ubuntu, and omitted from the version that comes with Windows? Someone did something wrong - I have done some rust development at work on CentOS/Ubuntu machines and all of those packages couldn't be more straightforward to install via yum/apt. That said it's not necessarily uncommon for some distributions (Ubuntu) to not ship with a complete development environment installed since there's a fair number of end users who may only be using stuff via snap or the Ubuntu software library or whatever. Some distributions differentiate between python2 and python3 since there's tons of legacy stuff still running 2. python3 (or python36 for f.e. version 3.6) is a pretty standard way of calling a specific version, but you can link the basic python command to whatever you want. Consider checking out update-alternatives in Ubuntu for a user-friendly way of handling multiple versions of the same binary. It can be normal to need to set the executable flag on things, depending on how you wound up with them. For example if you cloned it out of a git repository and the maintainer had it set as executable, it should get pulled down as executable; if you're just downloading stuff from wherever using wget or something then you may need to set it executable. Most Linux users will see something ending in .sh and know it's runnable though. I would suggest not including any such system dependency installation in build scripts and instead adding in the documentation what the dependencies are and let the end users make the call, at least as far as system packages go. Sheep fucked around with this message at 00:45 on Dec 19, 2018 |
# ? Dec 19, 2018 00:35 |
|
Dominoes posted:Thank you! I should clarify that Rust installed and updated without issue, but installing wasm-bindgen-cli with Cargo triggered the errors. The wasm-bindgen guide doesn't mention the dependency issues; it simply instructs to install with Cargo. I'm suspicious most linux-users would be able to immediately recognize how to solve all the errors I mentioned, or already have the right tools installed, but I'm trying to make this as frictionless as possible for a user of uknown experience. The issue I think is you have two different package managers which don't (can't) talk to each other: Cargo and apt-get. apt-get is the Debian/Ubuntu system package manager, which manages the kernel and all the system packages (pretty much everything under /bin, /lib, etc) and knows how all they depend on each other. Cargo is the Rust package manager which knows everything about rust packages and how those depend on each other. The disconnect is that Cargo may install packages that depend on packages managed by apt-get but since they're two separate package managers Cargo can't automatically install those dependencies. The reason Cargo can't just call apt-get is that you may be running it on a system with a different package manager (dnf on RedHat, pacman on Arch, brew on MacOS, etc.) and different system package managers packages also tend to have different names/naming conventions. You'll run into this issue pretty much any time you use a programming language-specific package manager to install a package with system dependencies.
|
# ? Dec 19, 2018 00:42 |
|
Python unfortunately comes in 2 major flavors, not terribly compatible with each other. Everyone should be on Python3 by now, but there's still lots of legacy stuff that relies on Python2. Larger distros like Fedora deal with this by having 2 separate dependency chains for everything (xxxx2 vs xxxx3, so there's python2 and python3, pip-2 and pip-3, etc. They also have another DEB/RPM you can install which sets one or the other to be the "default", so that calling "python" will choose either python2 or python3. And yeah, if you don't want your devs to deal with any of this then set up a script for them to run that does all the installation for you. I don't know much about WSL so I might be off base here, but personally I'd suggest using a headless VM instead of WSL. That way you're not restricted to whatever flavor of Linux WSL imposes on you, and you can pre-configure and supply VM images to your devs instead of them having to run through the same setup steps every time. The downside is that you'd also have to configure shared dirs between the Windows host and the subsystem if they wanted to use an IDE.
|
# ? Dec 19, 2018 00:53 |
|
Mr.Radar posted:The issue I think is you have two different package managers which don't (can't) talk to each other: Cargo and apt-get. apt-get is the Debian/Ubuntu system package manager, which manages the kernel and all the system packages (pretty much everything under /bin, /lib, etc) and knows how all they depend on each other. Cargo is the Rust package manager which knows everything about rust packages and how those depend on each other. The disconnect is that Cargo may install packages that depend on packages managed by apt-get but since they're two separate package managers Cargo can't automatically install those dependencies. The reason Cargo can't just call apt-get is that you may be running it on a system with a different package manager (dnf on RedHat, pacman on Arch, brew on MacOS, etc.) and different system package managers packages also tend to have different names/naming conventions. You'll run into this issue pretty much any time you use a programming language-specific package manager to install a package with system dependencies. I've seen and used one sort-of solution to this type of problem, http://cpanspec.sourceforge.net/, which will take a package from Perl's CPAN packaging system and generate an RPM .spec file you can use to build an .rpm for the Perl package. Unfortunately it's not perfect and doesn't always get the dependencies right, but you can always go and fix up the dependencies in the .spec manually if you find out that they're insufficient. Perhaps there's something like this for taking Cargo packages and generating .deb packages (or their source files)?
|
# ? Dec 19, 2018 03:15 |
|
Just finished setting up my in-home media server on Fedora today, a project I wanted to tackle all year but never could for reasons. Is the Wayland future really just going to deprecate gksudo without replacement? I used to have to turn in homework in vi so I guess I can survive, but I wouldn't tell the friend I pitched Ubuntu how to etc fstab in a terminal-based editor. Wayland's devs aren't wrong that running graphical programs as root is dangerous, but it seems very against the FOSS philosophy to not let the user do something horrible so long as they were warned first.
|
# ? Dec 19, 2018 04:05 |
|
Sheep posted:It can be normal to need to set the executable flag on things, depending on how you wound up with them. For example if you cloned it out of a git repository and the maintainer had it set as executable, it should get pulled down as executable; if you're just downloading stuff from wherever using wget or something then you may need to set it executable. Most Linux users will see something ending in .sh and know it's runnable though. Mr.Radar posted:The issue I think is you have two different package managers which don't (can't) talk to each other: Cargo and apt-get. apt-get is the Debian/Ubuntu system package manager, which manages the kernel and all the system packages (pretty much everything under /bin, /lib, etc) and knows how all they depend on each other. Cargo is the Rust package manager which knows everything about rust packages and how those depend on each other. The disconnect is that Cargo may install packages that depend on packages managed by apt-get but since they're two separate package managers Cargo can't automatically install those dependencies. The reason Cargo can't just call apt-get is that you may be running it on a system with a different package manager (dnf on RedHat, pacman on Arch, brew on MacOS, etc.) and different system package managers packages also tend to have different names/naming conventions. You'll run into this issue pretty much any time you use a programming language-specific package manager to install a package with system dependencies. minato posted:Python unfortunately comes in 2 major flavors, not terribly compatible with each other. Everyone should be on Python3 by now, but there's still lots of legacy stuff that relies on Python2. Larger distros like Fedora deal with this by having 2 separate dependency chains for everything (xxxx2 vs xxxx3, so there's python2 and python3, pip-2 and pip-3, etc. They also have another DEB/RPM you can install which sets one or the other to be the "default", so that calling "python" will choose either python2 or python3.
|
# ? Dec 19, 2018 20:10 |
|
Are there any desktop environments that play well with resolution scaling? Google searches tend to just point to people attempting to state that a 768p TN panel from 1999 is just as good as any modern IPS display and that hiDPI screens don't really matter. I use a 1080p T480 and 150% scaling is basically perfect - any smaller and this frankly lack-luster screen starts to become hard to look at. Windows 10 handles it well enough enough, but it seems like Linux in general hates anything other than pixel doubling.
|
# ? Dec 23, 2018 01:36 |
|
I'm using a T470 @ 1080p as I type this. I'm using the KDE spin of Fedora. It's fine. I don't have any special scaling percentage set: just how it comes when you install it. EDIT: Sorry. T570. It's getting late. I forgot what machine I'm using. apropos man fucked around with this message at 03:02 on Dec 23, 2018 |
# ? Dec 23, 2018 02:36 |
|
IuniusBrutus posted:Are there any desktop environments that play well with resolution scaling? Google searches tend to just point to people attempting to state that a 768p TN panel from 1999 is just as good as any modern IPS display and that hiDPI screens don't really matter. I've messed around with most DEs looking for a perfect fit for my 13" hiDPI screen on my XPS 13. Gnome 3 and KDE Plasma both work pretty well. Once I got rid of the annoying sidebar and installed a dock, I am pretty happy with how Gnome 3 works. In Fedora and Ubuntu, it automatically scales between 100% and 200% depending on whether I am using my 4k monitor or just displaying content natively on the laptop. If you are dead set on simple, Pantheon scales really well right out of the box but I am pretty sure it's tied to Elementary OS. Hekk fucked around with this message at 07:38 on Dec 23, 2018 |
# ? Dec 23, 2018 07:35 |
|
The key for HiDPI seems to be to running in wayland mode
|
# ? Dec 23, 2018 16:10 |
|
|
# ? May 2, 2024 12:52 |
|
Yeah, this is exactly what I want to do while visiting family. Will generate no friction whatsoever. (The story being linked to is basically just a new release announcement with some traditional WINDOZE SUX mixed in)
|
# ? Dec 23, 2018 16:31 |