Jonathan Perkin

· 3mo ago

First half of new shed floor built. Starting to regret not buying a first fix nail gun, hammering 100mm nails by hand isn't much fun, especially at the start when there's nothing to drive against.

#woodworking #diy

8

2

1

0

· 2mo ago

Decided to dig into what I initially suspected was hardware issues on a host performing "pkgin upgrade" very slowly. No, turns out it's legitimate requests hammering the disks, with pkg_delete performing an MD5 check on every file it is about to forcibly remove anyway. Eventually it'd be nice to add a flag for this, but for now I've removed it for my builds: https://github.com/TritonDataCenter/pkgsrc/commit/87055e3a0c53f409a0cf1efa887a3cb96afe0757 This will be in macOS and SmartOS builds soon, and will significantly improve upgrade performance. #pkgsrc

5

1

5

0

· 2mo ago

Almost watertight now, will be a lot less stressful once the rubber roof is on and given a chance to relax a bit, don't want wet OSB! Hopefully the rain will hold off for another couple of days.

Managed to fit most of the roof from inside, up a step ladder. Trying to avoid working from heights as much as possible.

Good fun though, and all done with basically a hammer, impact driver, and a circular saw. The tennis elbow is perhaps telling me that a nail gun might be worth it next time.

#diy

5

0

· 2mo ago

Replying to

@brubarwal@mastodon.sdf.org @ParadeGrotesque@mastodon.sdf.org Yeh it's a common issue, especially when you add CDNs into the mix. The only way for pkgin to work reliably is if entire package directories are only ever updated atomically. I do this for https://pkgsrc.smartos.org/ so you'll never see similar errors there, but I don't know that everyone else is so careful. If you just run naive rsync then files and pkg_summary will be out of sync until the entire directory has finished uploading. #netbsd #pkgsrc

pkgsrc.smartos.org

Binary packages for SmartOS/illumos, macOS, Linux, and NetBSD

Binary pkgsrc package sets for SmartOS/illumos, macOS, Linux, and NetBSD.

4

0

· 3mo ago

Fable dishing out the compliments that those of us running these systems for decades have always known and advocated for. "That DTrace trace was the whole game, by the way — owner identity from mdb plus lock/unlock pairing found one orphaned acquisition in 55k events and its exact stack. Nice platform to debug on." #illumos #dtrace #smartos

5

0

1

0

· 6mo ago

Yesterday we won parkrun bingo. It was my 100th, Natasha's 50th, also her 25th volunteer, and she brought along a friend who ran her 1st.

Oh and it was her birthday.

#parkrun #running

16

2

3

0

· 5mo ago

Volunteered to be the 26 minute pacer at parkrun this morning.

Mission accomplished.

Also enjoyed an absolutely stunning long run home along the North Downs Way, bluebells for miles.

#parkrun #running

14

0

3

0

· 5mo ago

First time parkwalking at parkrun today (screw you Nike!), as I have the Reverse London Marathon at midnight.

Beautiful bluebells still out around Priory Park.

#parkrun

12

0

1

0

@ParadeGrotesque@mastodon.sdf.org https://www.netbsd.org/cgi-bin/sendpr.cgi?gndb=netbsd

· 2mo ago

Replying to

netbsd.org

3

0

· 3mo ago

Looks like I picked the wrong week to quit sniffing^W^Worder a few tonnes of timber that needs transporting to the back garden.

We're currently experiencing a heatwave in the UK. I'm sure many of you would laugh at the "that sounds like a normal Tuesday to me" temperatures, but I am definitely not built for it. I sweat when I go running in minus celsius conditions.

(If you do not support Net Zero, or are a climate change denier, then you're an idiot and we can't be friends).

4

1

0

· 2mo ago

Replying to

@ParadeGrotesque@mastodon.sdf.org could you raise bugs for this please? it's quite frustrating as pkgin itself is fine but occasionally there are issues with the published package sets, especially around atomicity of updates.

2

1

0

· 5mo ago

If you've ever built a non-trivial amount of software on Solaris / illumos you've almost certainly hit math function errors like this: error: call of overloaded 'log(int)' is ambiguous I got fed up of adding patches to pkgsrc, and I believe I have a patch that fixes this once and for all. https://www.illumos.org/issues/15209#note-4 I'd appreciate wider testing. I've already tested it in a full ~28,000 package bulk build, but changes like this terrify me, and you can't be too careful. #solaris #illumos #pkgsrc

illumos.org

Bug #15209: Fix C++ math function visibility - illumos gate - illumos

Redmine

7

8

4

0

https://github.com/jperkin/bob/blob/main/examples/smartos-trunk.lua

· 6mo ago

Now that I've switched the daily SmartOS pkgsrc bulk builds fully over to bob, I figured it might be interesting to also publish the exact config file I'm using to do so.

No more complicated and fragile sandbox scripts, patches to pbulk, convoluted build infrastructure, etc.

Just a single binary installed using cargo, and a single configuration file.

Next release will be 0.99.0, we're almost ready for v1.0.

#pkgsrc #bob

8

6

0

· 5mo ago

Probably my favourite Half Marathon so far, and all because I finally paced it properly with a decent negative split! Run Gatwick was pretty cool.

#running

5

1

2

0

· 6mo ago

I realise the "blazing fast" meme is getting very old at this point, but it's genuinely so much fun making Rust code even faster. You often end up with simpler, cleaner, more idiomatic code too.

The original C pbulk presolve code takes:

real 0m35.686s
user 0m35.229s
sys 0m0.163s

Bob was already much faster:

real 0m2.115s
user 0m1.677s
sys 0m0.174s

But the new code is ridiculously (some might say blazing) fast:

real 0m0.506s
user 0m0.396s
sys 0m0.106s

#pkgsrc

7

1

0

https://gist.github.com/jperkin/72075147e330b35eb4869898207e828e

· 5mo ago

I wrote up a quick howto for getting bob up and running on OmniOS for someone, but figured it might be useful for others who want to play too.

Quite a minimal, stripped-back config, so doesn't have dynamic jobs configured. Easy to add though if you want to push your system to the max.

#pkgsrc #illumos #omnios

gist.github.com

3

2

3

0

https://mail-index.netbsd.org/pkgsrc-changes/2026/04/10/msg343920.html

· 6mo ago

In other news, boy am I glad I do not have to write python for a living.

Single missing comma causing completely different behaviour.

mail-index.netbsd.org

3

0

1

0

· 3mo ago

A hydration break 2 minutes after a lengthy stoppage where all the players had a drink.

America, you absolute joke.

1

0

https://mail-index.netbsd.org/pkgsrc-users/2026/03/25/msg042765.html

· 6mo ago

bob v0.9.0 now available!

I'm really proud of this one. Dynamic MAKE_JOBS and WRKOBJDIR in particular will save me a significant amount of time, and ensure my builds automatically stay optimal.

As always please let me know if you run into any issues or have suggestions for the next release.

#pkgsrc #bob #rustlang

mail-index.netbsd.org

3

0

3

0

· 5mo ago

Very disappointed to see Anthropic have dropped support for alternative platforms. I spent a lot of time using claude on SmartOS helping to do a lot of the grunt work debugging thorny issues.

Yet another reason why I cannot wait for viable local LLMs.

#llm #claude #illumos

2

0

1

0

· 5mo ago

I'm considering bumping the macOS requirements for my binary package repository available at https://pkgsrc.smartos.org/install-on-macos/ again.

There's already a bunch of packages missing because they have newer C++ requirements than Xcode 15.4 supports.

What OS are folks running?

#pkgsrc #macos

pkgsrc.smartos.org

Install on macOS

Binary pkgsrc package sets for macOS

2

6

5

0

https://reports.pkgci.org/Darwin/14.5/arm64/20260413T215911Z/report.html

· 5mo ago

Bob now automatically reports on, and links to, commits made to packages that fail to build since the previous attempt.

This is really helpful for showing regressions, and I hope will be useful to developers who want to check whether their updates caused problems on platforms they do not have access to.

#pkgsrc #bob

reports.pkgci.org

2

0

1

0

https://github.com/TritonDataCenter/pkgsrc/commit/2fb7d262707d40aa1db31599eaf9c0b952d15bd4

· 6mo ago

Currently testing a couple of Claude-inspired improvements to bmake and mksh, providing a decent speedup on macOS.

https://github.com/TritonDataCenter/pkgsrc/commit/3bd175884bb65350b24f442f8e3e9fcd45e695ce

Saw a 30% improvement spawning commands in a tight loop when using posix_spawn(), and avoids creating temporary /tmp/make* files for the vast majority of make commands. Both of which also help to reduce system time.

As a side benefit it greatly improves the readability of pstree output.

#pkgsrc #llm

2

0

· 12mo ago

pkgin 25.10.0 released.

Fixes an issue where changes to the pkgdb outside of pkgin operations (e.g. manually running pkg_add) could affect subsequent upgrades and lose autoremove information.

I'll get this into the SmartOS and macOS trunk builds over the weekend.

#pkgin #pkgsrc

7

0

2

0

· 6mo ago

Are you LLM curious, but don't feel you are getting the end results that other people claim?

Try starting your initial prompt with two simple words: "Let's discuss ..."

My experience is that it produces far better results. Clarify up-front what the goals are, deep dive into any technical aspects that affect codegen, and reason through any tradeoffs.

Basically the same as if you were asking a real person to do work for you.

2

1

0

https://whois.arin.net/rest/net/NET6-2630-2

· 34mo ago

I have many, many issues with #IPv6, but at least there was always the confidence that the address space was never going to run out, because nobody would be dumb enough to repeat the mistakes learned from IPv4 allocation, right?

Oh.

29

9

23

0

· 5mo ago

Replying to

@aru@functional.cafe oops, thanks, fixed. somehow mastodon removed the space before "again".

1

0

· 5mo ago

Oh FFS Mastodon. Poll completely wiped out because I made a tiny edit to the text portion of the post.

That's not only completely and utterly useless, there was absolutely no warning it was going to happen, and now I can't even see what the previous results were.

#mastodon

1

3

1

0

· 5mo ago

Replying to

@stefan@stefanbohacek.online If it behaved consistently I could perhaps see it as being obvious. But the fact that replies are retained means you could easily post something like "Reply with 'I agree' if you think xyz" and then just switch the meaning completely (and for many reasons that's arguably way worse).

1

0

· 5mo ago

Replying to

@wsb the repo currently supports 14, the poll is to see if I can bump that to 15 without breaking folks (looks like possibly not yet).

1

0

· 5mo ago

Replying to

@tux0r @zolaris I should also clarify that the daily builds I was referring to in the OP aren't the same as the ones that are published as binary packages. Similar, but different platforms and configs.

1

0

· 6mo ago

Replying to

What I particularly like is how things like gettext embed multiple identical copies of gnulib that all need patching: patch-gettext-runtime_gnulib-lib_localename-unsafe.c patch-gettext-runtime_intl_gnulib-lib_localename-unsafe.c patch-gettext-tools_gnulib-lib_localename-unsafe.c Good work Team C.

1

0

· 6mo ago

Replying to

@bentsukun is it possible to determine it automatically from the report URL? I was trying to improve the density of the report, and the machine URL is of no use to the majority of users. It's always going to be in the same directory and just be report.xz or report.zst instead of report.html

1

2

0

· 6mo ago

Replying to

@bentsukun let me know if we need to do anything to have this picked up and processed by bulktracker correctly.

1

3

0

https://github.com/ghostty-org/ghostty/discussions/11298

· 7mo ago

Upgraded to Ghostty 1.3.0, some nice improvements but didn't fix my biggest bug-bear, so have ended up raising a discussion about it.

Don't know if anyone else notices the same. Feels like such an obvious bug it makes me nervous that I'm doing something fundamentally wrong.

0

https://mail-index.netbsd.org/pkgsrc-users/2026/04/08/msg042818.html

· 6mo ago

bob v0.99.x now available, as well as a wip/bob package for testing.

These are release candidates in preparation for v1.0 which will be considered fully production ready.

If you have any questions, comments, or concerns, then now is the time to raise them as once v1.0 lands I will not be allowing any incompatible changes.

Happy building!

#pkgsrc #bob

mail-index.netbsd.org

0