No more fighting with a loose TTL-USB-cables: I have USB hub shield with USB-to-UART port :-) Or two of these: one for Raspberry Pi 3B+ and other for VisionFive 2 RISC-V SBC (in the pic). Need to still pile a TPM2 chip to the pins on top of the shield and hopefully it will still work. #arm #riscv #visionfive2 #raspberrypi

I.e. find IP of the keyserver with host/nslookup/ping and then pass it to --keyserver.

I quite often have to use IP for keyserver in #GnuPG's #macOS version. It does not otherwise find the route to the host. Never happens in any other platform...

next step in ai is obviously dynamically self-learning large language models. the current ai is google search with pre-computed aggregation heuristics, and nothing more than inspiring that.

not sure what to think about the next obvious step other than it is inevitably the next step :-)

so to scope the "problem" with the current AI is that it is not an "artist" using its own imagination, it is an "entertainer" following the patterns known to work for most.

@seindal @zwol I could imagine that at the time make + m4 was quickest way to PoC a higher level build system than raw makefile because you get away without having to implement a macro language so i'd guess that autotools was an unfortunate accident :-)

#pelican seems to be in sweet spot for me generating a site because it has first-class #reStructuredText support, i.e. no need to add any plugins it is built-in.

The benefit in it is that if I write anything with reStructuredText it is easy to port as kernel documentation later on. #python

@seindal @zwol reminded me of this presentation where the presenter totally mixes up with the toy example ;-) https://youtu.be/ULZxHSPWn98?t=1005

imho, m4 has its time and place in history and inspired bunch of other tools.

i remember using it ages ago for e.g. static web site generation where it did the job (today i'd probably pick something more modern), and of course still continuing enforced use with horrific autotools from time to time :-)

There's one killer feature in Serum that is pretty much non-existent in any other synth despite vast amount of plugins these days: render to osc. You can render one second of key press to wavetable. Then e.g. with something like #Polyend #Tracker you can use that wavetable to act as an filter and make effective use of wavetable mode given the limitations of the hardware. Serum and discoDSP Bliss 2 are essential plugins with hardware samplers and alike...

The first feature in #Kilohearts #PhasePlant that has made me turn into it more than #Serum is pretty cosmetic one: approachable way to modulate slope of an envelope.

E.g. with a bass patch for Serum I layer it with #Bitwig macro that has access to slopes through DAW interface but I cannot put it as part of preset. In Phase Plant I can have the macro within the preset.

#audio #musicproduction

@vbabka With TPM2, systemd-boot and FIDO2 logins you can pretty much set similar authority guarantees for machines at local premises, as you can do with SNP, TDX and SGX. Confidential computing is sort of like VPN but for computing, bringing attested remote resources. In appliances you could even do extra shielding by encrypting root and home partitions with different security policies.

Have been pretty happy how suddenly TPM2 has so much upbringing :-) When I implemented the support back in 2014 there was many years of with very few people paying any attention so this sudden interest has been totally unexpected. It is pretty de-motivating something that nobody wants to use.

@vbabka Something I just discovered anyway :-) I used Yubikey only for OpenPGP until very recently have started to use it for login into Google account etc.

Looking into pam_u2f.so for configuring #FIDO2 login,

Irritating thing #PayPal is that you can only add one hardware security key to your account.

@baptnz How does this connect to typst? I mean here the goal was to provide text paragraphs for typst document with arbitrary markup (not just links). I tried to use single bullet list also for the paragraph but that adds the bullet also to the main document (if I recall correctly). Just wondering if there is a simpler way to do same as I did, this was so far simplest way I found...

@liw If it is x86 I prefer Intel chipset, including GPU because Intel tends to enable Linux months before products hit the market. I.e. usually that kind of stack tends to be somewhat stress-free. This comes from ex-Intel employee but I think Intel has the best Linux team across the board when it comes to hardware companies :-)

@mpdesouza @monsieuricon @sj you can with "flatpak run" but yeah distrobox is definitely quick fix for this, thanks, had forgotten it :-) i mean it needs to be only "batch run" to update maildirs so not that inconvenient.

After couple of hours poking around I now know how to embed #typst markup from yaml to a typst document :-)

I’m working on resume made with typst (as I’m looking to find a new job by end of September), and in that I’m using #yaml file to separate presentation from the content. Job descriptions can involve links, and thus those entries need to be evaluated.

First observation was:

#eval(job.description, mode: "markup")

If mode-parameter is not defined, typst will try to parse the string as an expression (i.e. mode: "code" is the efault, which is feasible in this case.

Another problem was the use of the hash character for tags, which is used for comments in yaml.

I sorted that out by putting every description into double quotes after trying a few different approaches:

    description:
      "I first worked on Keystone Security Monitor for RISC-V by enabling it
      for CVA6 running on FPGA. This work was part of the EU funded
      #link(\"https://www.spirs-project.eu/\")[SPIRS project].
      It involved
      tuning the
      #link(\"https://buildroot.org/\")[BuildRoot]
      based embedded stack, and
      fixing various
      #link(\"https://github.com/keystone-enclave/keystone/issues/378\")[issues].
      in the OpenSBI firmware.
      For the second half of my contract I'm enabling Linux for the new SoC's
      developed by
      #link(\"https://sochub.fi/\")[SocHub project]."

The screenshot shows the end result.

Nothing too complicated but took some time to find working patterns so putting here as a #note for myself :-)

I've only recently discovered boot-time tracing using bootconfig and moved away from huge kernel command-line strings. This is the best tutorial on topic: https://www.youtube.com/watch?v=BjiC_a13e_k

@mpdesouza @sj @monsieuricon Yeah sure, I was not sure what to ask. I guess the thing to ask would be one dependency that at least suse does not seem to have: perl-Inline-C

OpenSUSE has perl-Inline and perl-Inline-Python packages only.

Flatpak package would be probably solution that would acceptable for most kernel developers because anyone can install it. The only glitch in that would be flatpak run prefix but that is not a big deal.

@mpdesouza @sj BTW, I went through all public inbox packages in OBS but none seem to work out of the box (tried locally both with osc and rpmbuild). I tried tune them various ways but without any mentionable success..

One working way would be probably to use OCI container, flatpak, appimage or something that containerizes all the perl shenanigans but gave up for the moment with this...