graywolf / is on Nostr

SRE at day, learning GNU Guile and Guix at night. New to Emacs, love it so far.

Last Notes

2024-08-08 19:20:38

Few notes made during implementation of #srfi64 for #guile turned into a blog post: https://wolfsden.cz/blog/post/state-of-srfi-64.html

2024-07-31 11:17:11

I am thinking about resurrecting my #pinephone . Does anyone use #sxmo as a daily driver? It does look *very* interesting.
The lack of password for unlock is a bit disturbing though.

2024-07-31 09:01:08

- reply

@npub1chp…l3fs If someone replaced "fascists" in that statement with some group you personally like, would you still feel the same about this topic?

2024-07-22 16:36:53

- reply

@npub1rqw…r3ud Looking forward to emacs 101 when masteringemacs turns into money printer :D

2024-07-17 15:35:02

Is there reading anyone could recommend regarding when does declaration order mater in #scheme (#guile)?
Why I am asking, after >4 hours of debugging I realized that cond-expand causes declaration order to not mater (== the code starts to work). I would really like to understand what is going on...

2024-07-16 21:02:19

- reply

@npub1zgl…rqd7 If I might ask, do you plan to stick to srfi-64? Or do you plan to explore alternative API designs?

2024-07-16 12:08:22

- reply

@npub1zgl…rqd7 Oh... I wonder whether it would be a good idea to not export it then.
Anyway, thanks for letting me know, time to write my own I guess. How hard can it be right?

2024-07-16 11:25:03

I do not understand how to use #test-error in #guile #srfi64. Why does this
(test-error 'bb
(throw 'ff))
pass? I am using the test-driver.scm from Guix (I would hope it is right since Guix uses it), and this is the result logged:
test-name: load*:
location: test.scm:150
source:
+ (test-error 'bb (throw 'ff))
expected-error:
+ bb
actual-error:
+ (ff)
result: PASS
Hm...

2024-07-15 15:11:53

Multiple times already I saw a code like (filter-map identity some-list). I wonder, is there a reason to prefer it over (filter identity some-list)? My understanding is that they should do the exact same thing, with one being shorter to write. Is that correct?
#guile #srfi #srfi1

2024-07-14 10:35:00

The feeling when you try to subscribe to a mailing list to report bug in #srfi #srfi64 specification and are told that
ERROR: Please enter a valid email address
Well fuck you, my address is valid and does comply with RFC 2822. Fix your validation logic.

2024-07-07 13:39:45

Definitely my most cryptic expression yet:
`(,@'#$extra-flags
,(symbol->string handle))
I totally did not miss the ' on the first attempt.
#guile #guix #gexp #scheme

2024-06-17 21:11:57

It would be great to be able to get running shell inside the #guix build sandbox. Or have the (optionally?) stages be cached, so the configure and build does not have to be re-run every time.
Debugging failing tests due to the permission limitations requires re-running the build over and over and can be quite time consuming (luckily my package was pretty small, so build is fast).
How do people usually debug this? Is there a smarter way than to just keep tweaking and doing full rebuild?

2024-06-11 13:46:20

#guile standard library is bit weird. On one hand we get amazing things like (ice-9 documentation), (texinfo reflection) and somewhat niche (and amazing) things like (system base lalr) and (ice-9 peg). On other hand, there is no built in json or base64 support.

2024-06-11 12:56:54

- reply

@npub1fzx…t6h3 Would be nice to have that choice :/ At work I will be able to pick between Windows (with their Total Recall) and Mac (with OpenAI integration). Both suck. One slightly less, because magit works at usable speeds.

2024-05-28 17:54:20

Surprising amount of procedures in (#guix #gexp) does not handle utf8 input. Combined with #guile 's approach of just replacing the utf8 characters with #\?, it is pretty annoying foot gun.
I am putting together a patch, hope I will make it before the core-updates merge, it kinda rebuilds a lot...
( ' ﹋ ' ; )

2024-05-21 15:39:49

- reply

@npub19am…qjn6 Great to hear that and thanks to everyone involved for all the work :)
Will there be a postmortem? It is always fun to read (and learn from) what went wrong.

2024-03-15 08:51:49

- reply

@npub1p5f…u6aj @npub19am…qjn6 This was quite interesting read, and the VM solution is fairly elegant.
I wonder whether it would be possible to add CPU, date, ... fields to the package record type and, iff set, automatically spawn VM for the build of that specific package.
For the "In rare cases, there’s software influenced by kernel details not controlled by the build daemon", another one I encountered is that currently python (Guix package) does not build when Guix runs on a foreign distro Alpine linux.