SRE at day, learning GNU Guile and Guix at night. New to Emacs, love it so far.
Public Key
npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx Profile Code
nprofile1qqsdfj5x2mq9j5rn0cgckadtrwr6uwygkmegdktm2j9a25swex6cu9spz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsn03c6h
Show more details
Published at
2026-07-03T20:28:40Z Event JSON
{
"id": "f43a9293be06eec12a0ecccfc3453b16748a84afb6567fdada8a741ba388913b" ,
"pubkey": "d4ca8656c05950737e118b75ab1b87ae3888b6f286d97b548bd5520ec9b58e16" ,
"created_at": 1783110520 ,
"kind": 0 ,
"tags": [
[
"proxy",
"https://fosstodon.org/users/graywolf",
"activitypub"
],
[
"client",
"Mostr",
"31990:6be38f8c63df7dbf84db7ec4a6e6fbbd8d19dca3b980efad18585c46f04b26f9:mostr",
"wss://relay.ditto.pub"
]
],
"content": "{\"name\":\"graywolf\",\"about\":\"SRE at day, learning GNU Guile and Guix at night. New to Emacs, love it so far.\",\"picture\":\"https://cdn.fosstodon.org/accounts/avatars/108/373/923/154/035/693/original/715cdc2562380c2c.png\",\"nip05\":\"[email protected] \",\"fields\":[[\"Homepage\",\"https://wolfsden.cz\"],[\"Blog\",\"https://wolfsden.cz/blog.html\"]]}" ,
"sig": "9212697b59c3b2b2dfdb9c7c32473d95d3be79792fa4e42b83916d9c35dfb0c2995bb9d8f6a8571d508ffbe77d8752d26c16e83884d708537e141bccd92e0d58"
}
Last Notes npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf #shepherd can run out of file descriptors during #guix boot. Fun Sunday. npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…dmr0 For some reason I still see "Looking for substitutes..." messages in the build log even with --no-substitutes. Is there a way to detect whether some "bad" substitute was downloaded or do I just need to go ahead and nuke my laptop? npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf How do you all prefer your syntax-rules? (define-syntax-rule (my-when cond body body* ...) (when cond body body* ...)) for precision or (define-syntax-rule (my-when cond body ...) (when cond body ...)) for readability? And what about documentation? Evaluate @nprofile…f8tp{body body* ...} ... or Evaluate @nprofile…f8tp{body ...} ... For me, just `body ...' reads better (both in code and in the manual, but `body body* ...' is more accurate, so I am on the fence about this. #guile #scheme npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf When loading modules, is the loading serialized or parallel? Especially when threads are involved? I really hope it is the former, otherwise I am out of ideas how to work around the bugs in our (rnrs records syntactic). Would anyone here know? #guile npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…mvm4 Hm, "This can be configured ... by a repository admin.". That is a shame. Thanks anyway :) npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…mvm4 Is there a way to get notified about pushes to some repository, pulling is fine. The best I can currently think of is git ls-remote in a loop, but that is wasteful on both ends, so if there is a better way to get the information I need, it would be great. npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf Bike shedding time! I am porting let-alist to #guile . Since there will be some differences, I thought I would name it let-alist*. So far so good. However, I will also change it to use let*, and I am unsure if that should affect the name. There are 3 options: let-alist* let*-alist let*-alist* The last is probably most precise, but it has a little too much * for my taste. I probably like the first the most, despite it being slightly misleading. Now, what does the internet think? #scheme npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…jrff よくできました! npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…33hd I am curious, do you feel like the reviewing got easier? npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf Does anyone know how to prevent shepherd from adding the timestamp into the log file? I am using the #:log-file to redirect the output, but the program itself logs the time, so my logs now look like 2025-03-31 23:34:52 2025-03-31 23:34:52 INFO: Anyway, check done Hm. #shepherd #gnushepherd npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…33hd In short (character limit :)): 1. Per-log file configuration. Some files I want to rotate more aggressively than others. 2. Ability to exclude files. Could be done by 1. and #:rotation-size-threshold +inf.0. 3. Timed-based #:expiry combined with size-based rotation is bit unorthodox. I need ability to express "keep the main log and 4 rotated versions". 4. Runtime discovery (== "logs to rotate are in /var/log/service.d/ and have .log suffix"). 1., 3. are a must, 4. would be nice. npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf I am torn what to do about log rotation on a server I am trying to convert to Guix right now. log-rotation-service-type is too inflexible, and rottlog-service-type will be removed soon. I guess I will just copy rottlog-service-type to my own channel, so that it survives the upstream removal. It will be bit annoying to diverge from Guix proper in this aspect, but do not really see other options here. #guix #rottlog #logrotate #logrotation npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf Yeah, I think (simple-service 'f-allow-http nftables-service-type (nftables-extension (rule "tcp dport 80 accept"))) is much better than my previous attempt. https://wolfsden.cz/blog/post/journey-to-an-extensible-nftables-service-type.html #guix npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…33hd My ideal timeline would look like this: 3.0.11 - #71262 and new GUILE_DEFAULT_CONVERSION_STRATEGY environment variable. 3.0.12 - Always default to UTF-8 port encoding, even in C locale. 3.0.13 - Change default conversion strategy. Or, this could be done all at once in 4.0 I guess. Is 4.0 planned for this decade, or do you find that unlikely? :) npub16n9gv4kqt9g8xls33d66kxu84cug3dhjsmvhk4yt64fqajd43ctqmslmtx graywolf @nprofile…33hd Thanks for the suggestions (and for reading it ^_^). The thing is that I like having the output in a single file (/var/log/messages), less places to look into. I would just like it to be prefixed differently. Maybe #:log-file with a symbol (#:log-file 'outerr-1)? Or new #:log-label key? Hm, can I use #:log-file "/var/log/messages" or will things blow up? I will look into Mailutils. I already have sendmail configured and working, so it seemed reasonable to just use it. :)