Nicolas T.

@avsm @mitchellh I think Linux' PR_SET_VMA_ANON_NAME prctl comes close?

macOS having memory tagging in mmap(2) caught my attention in @mitchellh's post on fixing ghostty's memory leaks. Looks like it's used heavily in macOS already, but I can't find a Linux arm64 equivalent anywhere

Full recording of this well-presented #CCC talk at media.ccc.de/v/39c3-from-si…

One of the coolest things I've seen in a while: "From Silicon to Darude Sand-storm: breaking famous synthesizer DSPs", reverse-engineering and emulating *the* synth that shaped late 90s/early 00s EDM. Bridging #hardware, #DSP, some #maths, #programming and, of course, #music.

@pizuricv Even though I hate the thought, maybe this means we want programming languages that cater to this idea/pattern, with no support for abstraction or generalization since some LLM can codegen or increase design complexity all the time anyway. Like, say, #golang (initially).

Funny, was using exactly the same argument in the office today

@strlen Back in the day we built Baardskeerder, an append-only B-tree'ish thing, to act as both the log and the database for Arakoon, a multi-Paxos key-value-store. Earlier version used traditional separation in log files and something BDB-like (TokyoCabinet).

Log stacking can happen when using Raft/Paxos to replicate DBs w/ their own WALs. Apache Kudu consolidates: the tablet WAL is also the Raft log. There have been experimental in other systems by disabling local WAL and using consensus log for recovery. Is there a generic solution?

Advice to young engineers: Read a lot of code. Several more times than you write. Ideally not the code internal to company. Go find where the state of the art is, study it. Having worked in the docker codebase early in my career gave me a lot of skill boost long-term.

@ezyang Fair. Then best I can offer you is gist.github.com/NicolasT/5779c…, assuming all scripts are Python files. The approach can be generalized to "any executable file", or whatever seems suitable.

@eikke But I don't wanna indent haha

When working a bug, it's pretty common to make a self contained Python script that reproduces the problem. It has always seemed like a shame to me that you can't just directly chuck said script into your test suite. Is there a test framework oriented around this idea?

UK online stores don't want me to buy things from them because I have no UK phone number (yet). Even Vodafone won't sell me a mobile subscription without providing them with a valid UK phone number. Get it?

Even Belgian media are reporting about @natfriedman's Vesuvius Challenge! standaard.be/cnt/dmf2023101…

@eurostar_help It being "more or less fair" can't be the goal of a change like this. @EurostarUK could learn a lot from how @KLM / @AirFrance managed a (successful) transition involving some very invasive changes in their loyalty program a couple of years ago.

@eikke The anniversary date has been changed to Oct 1st for everyone, and you now have 12 months to accrue status points. This might be more or less fair depending on your previous anniversary and points, but this aimed to satisfy everyone.

Hey @Eurostar! Not sure how the My Thalys World to Eurostar conversion went, but... I had Club level until 31/01/24 and amassed 3308 status points already (likely more, since last rides weren't processed yet). (1/2)

@technosophos Hah, I was there a week ago. Really liked their filter #coffee!

@henry3x @lodeba Ja natuurlijk. Bedoelde gewoon dat >=3% AER niet onredelijk lijkt? ECB geeft momenteel meer...

@eikke @lodeba Je hebt dan wel een stevig wisselkoers risico ...

@lodeba Sorry, daarop moet ik je het antwoord schuldig blijven (maar zou 't wel graag weten).

DMARC reports: an XML file in a ZIP file. MTA-STS/TLSRPT reports: a GZip'ed JSON. 🙄🙄🙄

@ibuildthecloud @ejc3 No need for full Raft (with its log and whatnot). Can be simpler, see e.g. the PaxosLease algorithm (with fixes vs. what's in the original paper, though).

@ejc3 In memory raft vs disk based. Leader election is ephemeral information.

Leader elections causes so many writes in k8s. This could really be improved. Is somebody working on this? Because most of the leader election crap could be saved in memory on the api servers and not written to disk. It requires raft probably, but the complexity is worth it.

@yminsky Could check with the GHC #Haskell folks: I wouldn't be surprised if they sometimes regret adding the LLVM backend. Codegen only beneficial for very specific code, LLVM version compatibility comes with quite a cost/difficulties, slow.

People constantly ask whether OCaml would be better off with an llvm back end, and I honestly don't know what the answer is, but it's interesting seeing Zig move in the opposite direction.