/tech/ - Techloligy

>>2806

This can't be serious. Happy little goyim gets allocated everything while the highest class still own cars and eat whatever they want.

Cunt's still pretty fat for being on a calorie ration.

At times I've had Raspberry Pis as my 'main PC.' I have no problem with this so long as they offer an embedded motherboard like the N100's or older J4XXX/J3XXX units (but ARM of course) - I like building PCs and those little hockey puck Chinese PCs ain't gonna cut it.

I've read some speculation about how Valve's new VR goggles might give a push to ARM desktop, because they are going to run SteamOS and also have an x86 translator, and it is possible that they will have a UEFI (or some other standardized) bootloader so that you can install whatever OS you want. All of this together might actually lead to at least some standardization amongst ARM desktops, or so they say hope.

>>2599

This is not true. I programmed, hosted and managed an imageboard for about 100 people for 3 years, and didn't have this problem.

>>2594

Building the community is the hardest thing. And make sure you have good moderation tools and a good lockdown contingency measure if shit hits the fan.

All of you are mere plebs compared to the supreme Emacs Lisp used to power goatse.world.

https://goatse.world/home

https://github.com/andrewmetzner/gapzip

>>2846

Is it optimised for eww?

>>1210

coming here and thinking you have some kind of advice to give is funny as fuck

>sicp is a meme

how is that? it conveys accurately the essence of computer science through descriptions of mathematical problems that emphasize the most important aspects like the towers of hanoi and fixed-point functions. it also establishes the relationship between program structure and process structure for example in the peano arithmetic example. i fail to see how this book is a meme if you've actually read it or watched the lectures on youtube. it's old but its concepts are timeless… it's like saying euclid's elements are a meme because it's old. it just makes no sense.

>>2430

To clarify, I am thinking of starting learning plain C at first, but running TempleOS.

>>1221

>Tables

As in precomputed hashes? Works fine if it hasn't been salted. Take wpa, simple enough hashing BUT it uses the access point details as part of the salt, so even if 2 APs have the same password their hashes of that password will be different.

>>2665

You don't need the source code if you think emacs is better.

>>2493

This is a cool program and I like it. Thanks anon.

>>2493

Bulk rename utility not good enough?

Props on actually making something though, I'll run it through the 98-11 gauntlet and let you know how it is.

>>2731

And the input:

#Last file has to be repeated at the end witouth the duration, because such is the will of the Omnissiah.
file '1.png'
duration 5
file '2.png'
duration 5
file '3.png'
duration 5
file '4.png'
duration 5
file '4.png'

You can crop h264 videos in ffmpeg with the following command:

ffmpeg -i inputfile.mp4 -bsf:v h264_metadata=crop_left=0:crop_right=0:crop_top=0:crop_bottom=0 -codec copy output.mp4

Just replace the zeroes with how many pixels you want to cut from a ”direction”. For example, the remove the baked in letterboxing of an 1920x816 video that was uploaded as 1920x1080, you need this:

ffmpeg -i inputfile.mp4 -bsf:v h264_metadata=crop_left=0:crop_right=0:crop_top=132:crop_bottom=132 -codec copy output.mp4

There are 264 useless black pixels, so you remove 132 from the top and also from the bottom.

This seems like the thread to ask:

>Have noisy dog neighbours

>Setup mic, record screen with date+time display

>Now have a few 100 gigs of footage to sort through

Is there any easy way to display the full audio waveform so I can isolate the barks? Or some automated thing that'll pull out clips based on audio levels? I'm video editing retarded, my only experience is simple things like Lossless cut.

>>2831

Anon described a setup with two laptops in a work-from-home area. One is used for work, belongs to anon's employer, and is closely surveilled. The other belongs to anon. Anon would like to use his own laptop as an advanced terminal to operate programs on his desktop, presumed to be located across the room. Anon is planning to dedicate an Ethernet link to this purpose.

Post >>2825 seemed to suggest concern about isolating the laptop from the Internet at large and confining "Internet risk" to the desktop. An isolated Ethernet link would achieve that, as long as the desktop is not configured to route IP, which the Linux kernel can do.

>>2832

I almost forgot that you will want gigabit Ethernet if you plan to run a Web browser over the network because modern browsers have massive faggotry problems with their display code and waste enormous amounts of X11 bandwidth. This also wastes crazy amounts of CPU cycles if X11 is tunneled over SSH.

If you want to share a screen or otherwise have X11 clients "floating" between the laptop and the desktop itself, VNC may be helpful. VNC can also be useful to "condense" wasteful use of X11.

>>2833

>I almost forgot that you will want gigabit Ethernet if you plan to run a Web browser over the network because modern browsers have massive faggotry problems with their display code and waste enormous amounts of X11 bandwidth.

Well, that will definitely thwart my plans of using a Thinkpad with a Pentium inside as an X terminal. Although maybe Ladybird will somehow completely degay the very concept of a modern browser to the point that it even solves this problem, but I shouldn't bet on that. On the other hand, my current plan is to switch over to emacs, and eww might work though, but I haven't managed to post on smug with it. Maybe I should read the manual though.

>This also wastes crazy amounts of CPU cycles if X11 is tunneled over SSH.

Is it wasting the host's or the terminal's CPU in this case?

>If you want to share a screen or otherwise have X11 clients "floating" between the laptop and the desktop itself, VNC may be helpful. VNC can also be useful to "condense" wasteful use of X11.

I'll look into that too, although I only plan to use the laptop when I cannot access the main screen due to it being used for work.

>>2834 (me)

Although now that I think about it, lynx can be used to post here if I am not mistaken, and maybe eat can play nicely with that. Then I could open a thread in eww one side of the screen to see the pictures, and also open the same thread in lynx via eat on the other side so that I can write a reply to a post if I want to. Of course, I should also start fucking around with emacs and read all the manuals and whatnot, but that is well outside of the scope of this thread.

>>2834

> >This also wastes crazy amounts of CPU cycles if X11 is tunneled over SSH.

> Is it wasting the host's or the terminal's CPU in this case?

Both. SSH runs on both boxes to encrypt/decrypt X11 traffic in the tunnel. What one box encrypts, the other has to decrypt.

> >If you want to share a screen or otherwise have X11 clients "floating" between the laptop and the desktop itself, VNC may be helpful. VNC can also be useful to "condense" wasteful use of X11.

> I'll look into that too, although I only plan to use the laptop when I cannot access the main screen due to it being used for work.

"Condensing" X11 abuse-faggotry using VNC would involve running a VNC client on the "terminal" box connected to the VNC side of an Xvnc server on the "host" box. In this configuration, the X11 traffic never leaves the "host" box and VNC only transfers "screen repaints" (and user inputs) over the network.

Using VNC also allows the "VNC desktop" to persist when no VNC client is connected, and clients can connect from different places, such as the laptop and the desktop itself, if you want to "carry" a browser session between the work-from-home desk and the regular computer desk.

>>2835

Looks like you're starting to climb the learning curve. Good.

>yubikey is closed source

>nitrokey decided to switch to rust due to all those nasty bugs related to memory safety that must have plagued them for a long time (never mind that rust has only one compiler, and that the language is in a constant state of flux)

Apparently the current best bet for hardware authentication is to get something from this bunch of Germans:

https://www.zeitcontrol.de/?lang=eng

>>2637

What if I just want a normal spreadsheet?

>>98

OP here, coming back with a little warning.

>Snappy Driver Installer

Don't use this. Development has ceased some time ago and it was handed over to some faggot who added malware to it. Thankfully, it's open source, so there's a clean fork to succeed it:

>Snappy Driver Installer Origin

>Windows

Finds, downloads and installs all the drivers you need and can't find. Open source.

https://www.glenn.delahoy.com/snappy-driver-installer-origin/

>ouch

>linux

tarball (de)compressor with sane command, arch linux user can install it with `yay -S ouch`

>TEncoder

>Windows

Open source video/audio converter/downloader that includes third-party tools such as MEncoder, MPlayer, Youtube-dl and FFMpeg. Its main draw is MEncoder (Linux, Mac, Windows), which can deal with stubborn proprietary codecs FFMpeg may struggle to convert (such as old Intel Indeo AVI videos).

Note: On the "Video and Audio Options" window, adding specific MEncoder/FFMpeg settings in the "Custom Options" fields requires a space at the start and end of the field to prevent errors when encoding.

https://www.fosshub.com/TEncoder-Video-Converter.html

>Windows Memory Cleaner

>Windows (XP/Server 2003 and older)

Portable RAM optimizer.

https://github.com/IgorMundstein/WinMemoryCleaner/releases

https://eldritchdata.neocities.org

contains a ton of ebooks on lockpicking, arms, survival, medicine, urban combat, privacy, opsec and all that based stuff. also really interesting articles

use uBlock, uMatrix or NoScript to disable JS, else it won't let you enter.

https://sizeof.cat/ I only got this one but it has cool random stuff in the links tab.

>>2813

Forgot this link.

https://kiwifarms.st/threads/nulls-tier-list-for-internet-services.129513/

https://igwiki.lyci.de/wiki/Main_Page

This is the current location of the install gentoo wiki. Not sure if it's going to die again since the cert is currently expired.

>>2633

Actually, I don't think these are available outside of the file, I haven't looked at this in years. Either way, it works, I'm pretty sure the #define works just like C and it replaces it in the file as it's preprocessed.

3-letter enums

Compile with -std=c11

#include <stdint.h>

#define ThreeLetterEnum(X, Y, Z) ( ((uint32_t)X) | (((uint32_t)Y) << 8) | (((uint32_t)Z) << 16) )

enum InputDevice {

INPUT_KEYBOARD = ThreeLetterEnum('K', 'B', 'D'),

INPUT_CONTROLLER = ThreeLetterEnum('C', 'T', 'L'),

INPUT_MOUSE = ThreeLetterEnum('M', 'O', 'U'),

};

alias fuckwebp='cd ~/Pictures/Memes/ && mogrify -format png *.webp && find . -name *.webp -type f -delete && cd ~

Doesn't work right all the time but it gets the job done well enough.

>>2613

>LISP OS

Not quite. Maybe "userspace" wasn't the right term since it would be based on a microkernel and everything is in userspace, but you know what I mean. Maybe "interface language" is a better term. Instead of more typical shell scripting, most things would be done in the language. I'm leaning towards smalltalk, but I've also decided to make a scheme with affine types. It's just intended to be a toy language right now, but if it turns out to not be crap, maybe it could be expanded into the interface language.

I know a lot about x86-64 and writing a driver (PCI and not), I also know my way around UEFI and ACPI though not everything like the fucking offset of the global lock in the FACS by heart, I gave up on my OS journey a while ago after writing a ~9k line core-microkernel in C++ (big mistake), which theoretically could have booted the init system and let it load drivers before finishing device enumeration and creating device threads for the driver processes, full featured with support for IO coordination and mapping the physical memory of a process into the virtual address space of a driver and hardened for *any sort of* allocation and freeing, that I had no idea how to even begin to unfuck but also because of fucking PE32+, I should have just done BIOS because gnu binutils supports PE16 or maybe just made it separate from the bootloader, I am not the most responsible software developer but I can write an abstraction or two and algorithms, if you want to work together post a contact

and nevermind because BIOS did a one-sector bootloader so no way it was using PE16, must have been nice back then not being anally raped by microsoft

Installing OpenBSD on a 600X Thinkpad from 1999 is not exactly straightforward, just as I've expected. Looks like the problem is that 256gb is too big for the disk partitioning tool of the i386 release. Or the SSD with the PATA adapter is just fucked, but it worked with FreeDOS when I last tried fucking around with this machine. Maybe the machine spirits are just a bit tired and I should try it tomorrow.

>found method to unlock the BIOS of the T42

http://asknotes.com/2018/09/04/removing-supervisor-password-svp-on-thinkpad-t42/

>it then goes to setup, but in user mod only

>after exiting setup it reboots but doesn't boot from the SSD, instead goes back to setup in user mode

>if I unplug it from power, then it wants the damned password again at boot, and I have to short the pins of that chip again

>but that again only lets me enter setup in user mode

Maybe I should try to get new batteries and then it would remember that I've shorted the pins, but that doesn't change the fact that it refuses to boot. I'm starting to remember why I shelved these machines for a year.

>>2768

You may have to install coreboot just to make it work.

It looks like a backup copy of the supervisor password is stored in the flash chip, which is where its getting it from when you don't short the pins.

You may have a different BIOS version from the one where shorting the pins works.

>>2769

I also has that other T42 that was shot to pieces, and I think the chip is still on there. Could swapping them work, or is that just a way to fuck everything up even more? I want to learn how to solder eventually, and this would be a nice practical project once I've learned enough to do it. Besides, I also want to try soldering the memory slot on the T30, although this whole learning how to solder thing is at least a year or two away.

>>2770

Swapping the chips may or may not work, but they're just memory. What matters is the data in them. If you can find where the password is stored, you might be able to recover it directly, or at least erase it reliably.

Soldering should not be needed here.

Welcome to the world of hardware hacking.

>XFS appears interesting, but I can't see what really makes it worth considering over ext4.

But you described it! It's an extent based system that does not rely on having to guess optimal predetermined inode sizes in advance. Also it has very good multi core scalability.

>>2528

This is insanely old information. A lot of the conventional wisdom about XFS is utterly outdated. Like how it's bad with many small files. The metadata structures have completely changed since. It's actually more performant than ext4 in many areas where it's supposed to be inferior. As for data loss, backups no matter what. I use borgbackup. An anecdote of course, but I use XFS on everything no problems.

>>2530

I can't tell you about Bcachefs itself except for pointing out it's too new. What I can tell you is complex COW files systems like ZFS/Btrfs/Bcachefs can give you very bad performance. I had a 1TB HDD with btrfs to store all my memes and porn. This meant many files in a folder, in some cases thousands. Btrfs was frustratingly slow at opening those folders, like tens of seconds long. I tried everything including reformatting to no avail. I migrated to an old laptop HDD (less performance) with XFS and the performance was night and day. Obviously it still not lightning fast because of the shitty hardware, but now it's only a few seconds for a usable open folder.

As much as I like btrfs raid1 (not really a raid) which allows you to throw bunch of drives of wherever sizes, whenever, and have it all work out, I have to warn about performance. Maybe keep it simple?

>>2531

> This is insanely old information. A lot of the conventional wisdom about XFS is utterly outdated. Like how it's bad with many small files. The metadata structures have completely changed since. It's actually more performant than ext4 in many areas where it's supposed to be inferior.

For me, performance is secondary - ext filesystems do not eat data and that is why I use them. Yes, I use data=journal, too.

Good enough that, until I got a new switch with longer hold-up time, power failures would cause an entry to appear in the system log reporting the network link going down. Yes, on PC hardware. Yes, the box would run out of power and crash milliseconds later. That log entry would be flushed out to disk (or at least the journal) in that tiny window between the switch losing power and the box also going down.

> As for data loss, backups no matter what. I use borgbackup.

The really important stuff gets archived on optical media here. There is still a very small possibility of the underlying HDD blocks going *poof* that no filesystem can entirely prevent.

> An anecdote of course, but I use XFS on everything no problems.

OK then, do the test:

(for i in {1..50}; do cp -a /usr /mnt/xfs/usr-$i & done)

Then wait for 30 seconds with the disk light on (if the disk light isn't on steadily, start more copies by changing "{1..50}" to "{51..100}" or similar) and pull the plug. Reboot and see what you have in /mnt/xfs or whatever other XFS filesystem you used for the experiment.

Complete files are expected; they were copied before power was cut. Missing files are fine; the copy probably didn't reach them. Truncated files (at most one per copy) are expected; the data transfer was in progress when power was cut. Zero-length files are probably bad; the data was lost, OR (if it is the truncated file in that copy) the file was createPost too long. Click here to view the full text.

Can you recommend filesystems reading material?

I only know about "Practical File System Design: The Be File System"

>>2532

> [challenging >>2531 to test XFS]

Did he died?

ext5 when?

>>2156

>HAMMER2

https://www.dragonflybsd.org/hammer/

>HAMMER retains a fine-grained history. The state of the filesystem can be accessed live on 30-60 second boundaries without having to make explicit snapshots, up to a configurable fine-grained retention time.

>Coarse-grained history is controlled by snapshots. By default the system cron generates one snapshot a day and retains 60 days worth. Snapshots can be accessed live.

I am not sure how useful that is in real life, but it does sound impressive.

>>2341

OpenBSD has this site with all the currently available ports:

https://openports.pl/

I was thinking about switching over to OpenBSD, but two programs I use a lot are Librewolf and Freetube. Librewolf did have an OpenBSD port, but then they stopped developing that because fuck having nice things. Now, we are talking about *NIX, how have would it be to run the Linux version. Well, OpenBSD developers completely removed compatibility due to security risks. So be it, but can I run the FreeBSD version? No, fuck you, that part of the system got removed even before Linux. Really, they should make a fork called ClosedBSD, which is just OpenBSD but with those two brought back. ClosedBSD as in ''with closed eyes you can ignore the horrible security risks that come with running FOSS software from other FOSS OSes.

>>2491

Although it bothers me so much that I will still try to do my best to seek out available alternatives. For a browser I could do the cool thing and use surf with tabbed from suckless. And for youtube I could just follow the channels with RSS feed and use yt-dlp or mpv to watch them. The only problem is that my internet is not really up to the task of streaming, but I guess I'll just download everything and then delete them once I am finished. That could even help me keep in check how many tabs I have open in my browser, because a lot of them are just random videos I intend to watch later, and if I have already have them downloaded then I don't need to keep a tab on them. Who knows, maybe it will be fantastic.

https://vez.mrsk.me/freebsd-defaults

An interesting read about FreeBSD's (lack of) security. Long story short, use the OpenBSD equivalent of everything network related, and also the defaults from that article. I wonder if the same problems apply to DragonFlyBSD too.

I've heard good things about GhostBSD lately.

>>2757

> What other folders should I have on the SSD in addition to /boot? What folders should go on what partitions?

The basic principle is to tilt the SSD as far towards a read-mostly workload as you can to maximize its service life. In turn, use a good PSU and provide plenty of airflow around the HDD to maximize its reliability.

For filesystems, use ext4 all around, except that /boot/efi must be FAT so the firmware can read it. Since /boot and /boot/efi are so rarely accessed and your SSD is quite small, they should go on the spinning rust. For HDD-based filesystems, you should use tune2fs -o journal_data to route all writes through the journal, but this doubles write volume and therefore should not be used on SSDs, only HDDs which have "unlimited" write endurance.

You probably want the system on the SSD if it fits (128GB might be a bit small). Really "system" means /usr because /boot is only accessed when updating or starting the machine. Fetching the kernel from spinning rust is not going to make a significant delay and Linux never needs its kernel image from disk while running.

Swap needs to go on rotating media - swap on flash is fast but it will also kill your flash way too quickly.

The traditional recommendation was 10x RAM for swap, but with modern large RAM, 2x - 5x is more appropriate, especially since you can have multiple swap areas with Linux if you ever need even more space. I recognize that 2x RAM for swap is a quarter of your spinning rust. You might go down to 384GiB swap (1.5x RAM) or even lower, especially if you are sure that you will never use software suspend. Also, I think software suspend requires swap to be an actual partition because the kernel needs to resume before it would load LVM.

You should definitely put /home and /var on the spinning disk. Both of those are frequently written.

Traditionally, /boot was a mountpoint due to PC BIOS limitations. Now, /boot/efi must be a mountpoint due to UEFI limitations and /boot is typicPost too long. Click here to view the full text.

>>2757

>>2758

>Do not use btrfs - it has a well-earned reputation for eating data.

Either you're a 2014年から時空旅人, tried to run a RAID5/6 for some bizarre reason or Linus Torvalds on another imaginary mailing list rant with Kent Overstreet and accidentally typed Btrfs instead of bcachefs.

Having used Btrfs as my root filesystem for 8 years now on Hards and Solids, in time I've had about a dozen events where Btrfs would instantly screech after detecting misaligned free space for some reason and refuse to mount or mount as read only, necessitating the use of Rescue Media to get it back in a working state.

Despite the warnings on the Arch wiki and from Btrfs' own repair utility this has so far gone off without a hitch and I've yet to lose a single file, but the ext4 partition of my HDD did suffer the loss of a small number of reaction images and god knows what else after a power outage.

It refused to mount after that and I ran fsck which reported erroneous free space, after repair my folders were all there but those few files had broken metadata and couldn't be opened in spite of fsck reporting an all-OK on the partition.

Aside from that it hasn't had a single hiccup since then unlike Btrfs and the files gradually restored themselves through unknown means over the years, with the final one coming back just a few weeks ago and is much faster owing to its architecture.

What I'm trying to say here is that Btrfs is good if you have autism and care about data integrity at the expense of performance, while ext4 "just werks" but is ultimately vulnerable to bitrot like all plain journaling filesystems.

Not sure what Anon has against XFS though, it's a solid alternative to ext4 especially on HDDs where it still has a slight performance advantage but right now lacks online metadata checking which will be merged in Kernel 6.18, I use it on another drive with no known data loss since the partition's creation 5 years ago.

That sPost too long. Click here to view the full text.

>>2759

> I've had about a dozen events where Btrfs would instantly screech after detecting misaligned free space for some reason and refuse to mount or mount as read only, necessitating the use of Rescue Media to get it back in a working state

You realize that every one of those incidents represents the btrfs driver nibbling on your data and catastrophic destruction was only averted because its slow and extensive checks caught its own fuckup in time?

In contrast, the ext4 driver doesn't fuck up your data in the first place.

> the ext4 partition of my HDD did suffer the loss of a small number of reaction images and god knows what else after a power outage

That's what tune2fs -o journal_data is for. Since that's a bad idea on a SSD, minimize the writes to those partitions.

Incidentally, the greatest benefit you can get from an SSD on Linux is from putting the userspace executables on it, and those are only written when installing updates and new packages.

Linux treats executables as read-only swap for the program text of processes running from them. The zero-seek-latency of SSDs is beneficial here.

> Not sure what Anon has against XFS though

Lack of Power Fail Interrupt on PC hardware. You got bitten by this even with ext4. Under the same circumstance, XFS could have had catastrophic damage. garbage written to logical journal propagates widely upon journal recovery

Ext3 avoided this by using a physical journal instead of a logical journal, such that garbage can at most trash the just-to-be-written disk blocks after recovery. Ext4 adds support for checksums in the journal which also prevent even that damage. Has XFS finally also implemented journal checksums?

>>2761

> I'm planning on going with 1x RAM for my swap partition.

In that case, since you're using Gentoo, I suggest 1G /boot/efi, 3G /boot, 256GiB swap, and then the rest of your filesystems. If separating /home and /var, try 50G /, 300G /var, balance of disk /home, with the entire SSD assigned to /usr. Gentoo is a lived-in dev environment, so it should have bigger boot partitions than "packaged" systems need.

Given the small size of your SSD and use of Gentoo, you might want also allocate a "shadow" partition on the HDD matching the SSD size and keep a warm backup of /usr there. SSD failure? One small tweak to /etc/fstab from a rescue environment and you're back up. HDD failure? Well, the box is fucked anyway. Got tape? Me neither.

> I don't plan on ever using suspend. I will probably never need to swap anything.

Linux likes to steadily trickle pages out to swap "just in case" it ever needs to reclaim a page frame quickly. This is one of the reasons swap-on-SSD is a Really Bad Idea.

You might want to consider encrypted swap under a random volatile session key. No key management to worry about and the contents of swap are unrecoverable after shutdown, whether orderly or otherwise.

> I will probably just put /usr on the SDD, and have /boot, /boot/efi, and / on the HDD. Not going to do a separate /home or /var partition.

That's even simpler and probably should work.

I'd still advocate keeping a separate /home just in case you ever miss too many updates and have to reinstall Gentoo.

The classic rationale for a separate /var filesystem is that /var is one of the most heavily written filesystems and thus is most at risk for corruption, which is of course very bad if if hits your root filesystem. I've never had /var corrupted, so maybe improvements had made that rationale obsolete when I learned it.

&Post too long. Click here to view the full text.