Simon Martin's website - Simon's blog

Status Report - Week 18 2026

Unknown — Mon, 04 May 2026 00:00:00 +0000

Real life
Trust me: the false prophet</a> TV series, that I found very disturbing. On a lighter note, we also watched The Dark Knight</a>, that I really like, and started watching the Peaky Blinders</a> TV series - it's a re-watch for me, but I have forgotten almost everything so it's cool :-)</p>
I have finally finished reading Fundamentals of Physics: Mechanics, Relativity, and Thermodynamics</a>, and immediately bought the second volume; I wish I'd read it during my studies, because chances are that I'd have loved physics more... I also started the second volume of La Guerre et la Paix</a>, as well as Rhinocéros</a>.</p>
The week was quiet both in terms of running - only 23.33 km - and GeoGuessr</a> - only 7 games (4 wins), with a finish at 17th in my group.</p>

Open source </a> </h1>
I slowly got back to working on SourceHut</a>, and worked on a new revision</a> of my patch to process hub webhooks in Go instead of Python. It depends on minor fixes (more precisely, the same one in several services) that I encountered along the way and sent patches for: 1</a>, 2</a>, 3</a> and 4</a>.</p>
I also made the April release</a> of sr.ht-integration-testing</a>, that brings a big coverage increase around the project hub, and some daily test runner stability fixes that are working pretty well. We had only one failure due to flakiness this month, while we had 14, 9 and 3 in January, February and March respectively...</p>

Status Report - Week 17 2026

Unknown — Wed, 29 Apr 2026 00:00:00 +0000

Real life
DGSE: Une brève histoire du renseignement français</a>, that was interesting.</p>
I barely ran (5.16 km...) but we walked a lot (44.47 km) so it was an OK week sport wise. And I decided to take a short vacation at GeoGuessr</a> - we'll see if it makes me better, worse or has 0 effect.</p>

Open source </a> </h1>
SourceHut</a> regression with a trivial service patch</a> and a related fix</a> in hut</a>.</p>

Status Report - Week 16 2026

Unknown — Tue, 21 Apr 2026 00:00:00 +0000

Real life
Lyla, Nathalie and I watched Mulholland Drive</a> and I loved it even more than I did when I first saw it 25 years ago. And seeing Lyla go through various phases of WTF throughout the movie was fun :-)</p>
I finished American Sniper</a>; this is one of those meh books that you forget about as soon as you finish them... I'm almost through Fundamentals of Physics: Mechanics, Relativity, and Thermodynamics</a>, that continues to be excellent and teach me a lot.</p>
It was another mediocre week in terms of running, with 21.64 km, as well as in GeoGuessr</a>; I won 9 out of my 19 duels, enough to finish 18th in my division and stay in Master IV for another week.</p>

Open source </a> </h1>
This week was again slow on the SourceHut</a> front because the staff seems still pretty busy fighting the recent DDoS attack, and I don't want to stack too many patches on top of unreviewed ones.</p>
So I only finished and submitted the patch</a> to migrate existing hub webhooks to Go and prepared the hut</a> change to hook up the new GQL write operations in a branch</a>.</p>
I also sent small patches to sr.ht-container-compose</a>, to silence hut warnings</a> and setup another user</a>.</p>
The bulk of my time was spent on servo</a>, more specifically on the WebGL side of the house and flaky tests: those</a> three</a> patches</a> improve TexImage3D and should stabilize tests, while this one</a> fixes broken links in comments.</p>
Even though it leads to very small changes, I like this kind of work: you don't know anything, start pulling a string, and pull, and pull, and you learn "a lot" along them way - time well spent! :-)</p>

Status Report - Week 15 2026

Unknown — Tue, 14 Apr 2026 00:00:00 +0000

Real life
France Identité</a>. You need to be in front of an "official person" to set this up, and the plan was for her to do so while at the consulate. The idea is that you get an email with a personal link that opens an app on your phone, and you do some operations with that person. This app is coded so well that if you open it like any other app instead of via the personal link, it's "really bad", and you have to start over.</p>
It's apparently so bad that they warned her about this multiple times beforehand, and managed to scare her so much that she decided to only download the app while at the consulate. This makes a lot of sense, but the grave mistake she made was to assume that she'd have connectivity there. I almost suspect they have some kind of jammer because she barely had any signal - signing on the app store took minutes, downloading was super slow, etc. So much that after a long time, the tech person asked her to try outside the building, and if it did not work, to make another appointment. Needless to say, she was able to do everything in seconds from outside.</p>
So everything worked in the end, but it took a lot of time, the tech person was pissed off because what should take 5 minutes took almost 1 hour, and Nathalie was stressed out. All this because the guys who develop the app are so bad that not starting it from the personal link apparently "breaks everything". Beyond the fact that the development probably cost a million euros to France, I'm amazed that nobody</strong> thought that the UX was just unacceptable and that this should never go to Production. I guess the French administration created their own version of Amazon's "Insist on the highest standards": "Insist on the lowest standards".</p>
On the reading front, I continue Fundamentals of Physics: Mechanics, Relativity, and Thermodynamics</a>, still very good, and I started reading American Sniper</a>.</p>
I ran a bit - 22.84 km - but not enough to my liking. I need to correct this.</p>
Finally, I had an OK week at GeoGuessr</a>, with 21 wins out the 39 duels I played. This got me to finish 17th in my group, and stay in Master IV for one more week :-)</p>

Open source </a> </h1>
I made some progress on SourceHut</a>'s project hub, in particular the migration of webhook processing from Python to Go. I submitted two</a> more</a> pre-requisites to process all new</strong> webhooks in Go, via this patch</a>.</p>
I also drafted</a> the migration of existing webhooks, so that we can simply remove the Python code. I plan to spend a bit of time on testing and submit it next week.</p>
I will slow down a bit on "core" SourceHut patches because the staff is busy fighting a DDos attack</a> and there's not much time stacking too many PRs on still not reviewed ones.</p>
Instead, I plan to spend the time on increasing sr.ht-integation-testing</a>'s test coverage, probably adding support for scenarios involving multiple users.</p>
Finally, I checked if/how I could test one of my pending servo</a> patches</a>. It's unfortunately not possible to do it, so the patch got merged as is.</p>

Status Report - Week 14 2026

Unknown — Tue, 07 Apr 2026 00:00:00 +0000

Real life
On the reading front, I finished Wild</a>, and continued Fundamentals of Physics: Mechanics, Relativity, and Thermodynamics</a>. This book is excellent, with crystal clear explanations, and I'm really glad to have (a bit randomly) picked it up. It's a big one though, that I did not want to carry in the train to Dordrecht, so I've also started Russian Literature: A Very Short Introduction</a>.</p>
The motivation to go out running is still nowhere to be found, and I logged a mere 22.82 km; let's hope spring fixes this.</p>
Finally, I did not play much GeoGuessr</a>, with 9 wins out of only 17 games. This was enough for me to finish 15th in my division, and stay in Master IV.</p>
Open source </a> </h1>
SourceHut</a> project hub from Python to Go. It's a lot of boring code to write, and most of the time was spent on continuing to increase the coverage of the associated tests in sr.ht-integration-testing</a>.</p>
After I added tests</a> to cover all the "TODOs" I had left in the Go code as I was writing it, I am now confident that the Go code is good enough for submission. I plan to do it as soon as the current DDoS attack</a> on SourceHut is mitigated.</p>
This work depends on this patch</a> that I was able to submit before the DDoS started.</p>

Status Report - Week 13 2026

Unknown — Mon, 30 Mar 2026 00:00:00 +0000

Real life
I continued to read Wild</a>, and started in parallel Fundamentals of Physics: Mechanics, Relativity, and Thermodynamics</a>. I've always been more a maths/computer guy, and I never understood physics (which did not prevent me from being very good at it lol), and figured it was worth trying again after 20+ years :-)</p>
Otherwise, I ran a mere 18.07 km (IOW the "running fire" is not back yet :-)), and had a good week at GeoGuessr</a> with 22 wins out of 43, enough to stay in the Master IV division.</p>
Open source
SourceHut</a>'s project hub webhooks from Python to Go (the next step in my ongoing work about hub's GraphQL API), and quickly realized that it would be too dangerous without at least a few tests to catch possible regressions :-)</p>
I therefore</p>

resurrected</a> some sr.ht-integration-testing</a> tests that were based on UI parts recently dismantled, and changed them to instead query the event</code> database table, and</li>
improved them</a> to not only assert the number and structure of events, but also their full contents.</li> </ul> This is not fun and pretty tedious work, but it will drastically increase my confidence in my port of webhook processing from Python to Go, exactly what I want (and need).</p> And it already started providing value, since it helped me find an issue in version 4</a> of my patch to add write APIs (that I fixed here</a>).</p>
Status Report - Week 12 2026 Unknown — Wed, 25 Mar 2026 00:00:00 +0000 Real life</a> </h1> It's been a pretty standard week, except for the fact that our (brand new) dishwasher caused problems again. We bought it in November 2025, and had to call in a technician at the end of January because the I-automatically-open-towards-the-end-so-that-things-dry-better door would not close (everything else works - you just need to block the door with a chair :-)). He fixed it, but it broke again, so we'll have another technician and I'll push to get a replacement. This is completely unacceptable, all the more for an 800 euro machine! Siemens is definitely not scoring points in our household...</p> I finished The Heart of the Buddha's Teaching</a> and really loved it; you can expect me to mention more books about buddhism here in the future :-) I started reading Wild: From Lost to Found on the Pacific Crest Trail</a>, and in parallel resumed (after 8-ish years) Programming Rust: Fast, Safe Systems Development</a>.</p> I had a slow week in terms of running, with only 15.72 km - I'll try to do more next week.</p> Finally, it was my first week in the Master IV division at GeoGuessr</a>. Master IV and above work a bit differently: all the games count, you're only matched against folks with a similar level as you, and the end of week ranking is simply by ELO. I did not really like how "lower divisions" worked, so I'm happy, it's harder but fairer. I had a rollercoaster Sunday, losing a lot of games (with my ELO dropping below 550...) before I finally got my bearings back. I eventually finished at 713 ELO, ranked 19th of my group, enough to stay in Master IV \o/ Overall, I played 74 duels this week, winning 41.</p> Open source</a> </h1> I continued to do a bit of work on SourceHut</a>:</p> I submitted a fourth iteration</a> to add write operations to the project hub's GraphQL API.</li> I slowly started to look at migrating hub's webhook processing from Python to Go, and along the way identified a bug with repository deletion that I fixed</a>.</li> I submitted a new revision</a> to hook up the hub's read GraphQL APIs in hut</a>.</li> </ul> I also spent a bit of time working on servo</a>:</p> I fixed</a> an annoying display issue</a> in servoshell</a></li> I spent some time investigating an issue</a> with SVG scaling. I'm getting nowhere at the moment, but it's a good opportunity to learn about a domain I know nothing about</li> I submitted a patch</a> to fix a very minor WebGL compliance issue</a></li> </ul> Status Report - Week 11 2026 Unknown — Tue, 17 Mar 2026 00:00:00 +0000 Real life</a> </h1> It's been a pretty busy week, with lots of sport (mostly in front of the TV ;-)), quite a bit of coding and finally some progress at GeoGuessr!</p> This week was the last of the 6 Nations</a> rugby competition, that brought a lot of joy, stress, and a semi-happy ending: Ireland solidly won against Scotland (what I hoped for), Wales were excellent against Italy and won their first game in the last 3 editions (yaay!), and France played really badly against England (who were very gool) but eventually won, thanks to a penalty at the very last minute! I am really concerned that the French crumble as soon as they're facing some kind of serious competition: there will be lots of it in the World Cup next year...</p> On Sunday, Lyla and I ran the 10km race at the CPC Loop Den Haag</a>, in respectively 1h02:39 and 1h02:44 - that's better than expected given our limited amount of training this year. This got me to close the week with 18.75 km of running.</p> I continued reading The Heart of the Buddha's Teaching</a>, and finally made progress at GeoGuessr</a>: I won 11 of my 20 counting games, finishing 4th in my division, enough to be promoted to Master IV - at last!!! It's likely that I get destroyed there next week and immediately get demoted back to Gold I, but it's still a good achievement.</p> Open source</a> </h1> As usual, I worked a bit on SourceHut</a>, iterating</a> on the effort to add write GQL APIs to the project hub. I had to dive a bit into Go GQL clients generation, adding support for field aliases</a> (design</a>).</p> I also sent a revision of the patch</a> to keep track of superseding patch versions, and improved</a> hub project access via hut</a>.</p> I also fixed</a> a freeze in aerc</a> when quickly changing directories, and sent my first (trivial) patch</a> to the servo</a> web browser since... 8 years!</p> Status Report - Week 10 2026 Unknown — Thu, 12 Mar 2026 00:00:00 +0000 Real life</a> </h1> After a great ski vacation, we're back in our regular routine: kids are back to school, Nathalie gives tons of classes, and I keep myself busy with lots of random small things at home.</p> The 6 Nations</a> rugby competition is back as well after a one week break, and it has been an emotional roller-coaster. It started pretty poorly, with Wales almost winning against Ireland (Rhys Carre's try</a> was by far the best moment in the game), and Scotland completely destroying France - I'm afraid, they showed they won't win the World Cup next year, not for lack of skills but for their inability to operate when not everything goes their way :'( The weekend was fortunately saved by Italy (not) surprisingly winning against England [1], for the first time ever in history.</p> Next weekend will be the last in the competition and will be very interesting. I do hope that France wins (their game and the overall competition), because I am not ready for yet another French rugby drama :-)</p> I wrote last week that the last book</a> I read piqued my interest in knowing more about buddhism, so I started reading The Heart of the Buddha's Teaching</a>, and I like a lot so far.</p> I had a good running week, with 35.45 km, and I barely played any GeoGuessr</a> - just my 20 "counting games" on Sunday (9 wins). I surprisingly got my best Gold I ranking ever, finishing 8th in my group.</p> [1] This</a> works for me as well ;-)</p> Open source</a> </h1> I continued iterating on the GraphQL API I'm adding to SourceHut</a>'s project hub, looking at write operations after read ones landed</a>. I submitted this patch</a> along with this testing document</a>; I got some good feedback on the former, and will submit an updated revision soon.</p> I also submitted a patch</a> to keep track of the revision a patchset supersedes, if any, and expose it in the UI.</p> Things are quiet on the sr.ht-integration-testing</a> front (this will change once I'm done with the project hub's GraphQL API, that will need some not-manual-testing ;-)): I only increased the daily tester's reliability by automatically reclaiming disk space</a>) on the EC2 runner when needed (instead of erroring out stating that "someone needs to do something").</p> Status Report - Week 9 2026 Unknown — Mon, 02 Mar 2026 00:00:00 +0000 Real life</a> </h1> We had an AWESOME week in Alpe d'Huez, enjoying a lot of snow and sun - we probably saw 3 clouds in the whole week...</p> It's the first year none of the kids is attending ski school, which removes time constraints and brings a lot of freedom. It was really great to ski as a family, and enjoy quieter times on the slope (e.g. during lunch break).</p> Here are a few pictures I took this week:</p> From the top of Sarenne, at 3330m (notice the quantity of snow!) </p> </p> After the first wall on Sarenne </p> A pretty nice place to have a picnic </p> The kids leading the charge down the "La Balme" slope </p> The last picture this year... </p> In terms of reading, I finished reading The art of happiness</a> - a very good read that makes me want to explore buddhism a bit more.</p> I did not run, but (obviously) skied quite a bit - 219.26 km (!). And I decided to take a break at GeoGuessr</a>, and not do anything but play the Daily Duel to keep my streak.</p> Open source</a> </h1> No activity this week :-)</p> Status Report - Week 8 2026 Unknown — Mon, 02 Mar 2026 00:00:00 +0000 Real life</a> </h1> This is the first of two weeks of winter vacation \o/</p> We went to my mum's next to Lyon on Wednesday, and then to Alpe d'Huez on Saturday to have a week of skiing. It snowed a lot</strong> there this week, so we'll have a lot of snow, and the weather is expected to be very nice in "our week"; we're lucky.</p> I continued to read The art of happiness</a>, ran very little (6.75 km), and had a reasonably good week at GeoGuessr</a>, with 9 wins out of 20 games. I finished 13th in my division, well enough to remain in Gold I.</p> Open source</a> </h1> The only thing I did this week is submit a fix</a> for a regression in aerc</a>.</p> Status Report - Week 7 2026 Unknown — Tue, 17 Feb 2026 00:00:00 +0000 Real life</a> </h1> I must admit that I spent a lot</strong> of time this week watching TV, courtesy of the winter Olympics. I don't care about most events, but I am a big biathlon and cross-country ski fan, both sports have a lot of competitions, and the French teams are doing very well do far - I'm very happy.</p> During quiet times, I read quite a bit, and finished Overcoming the rating game</a>, and started-and-finished Un avion pour les gouverner tous</a> and Pour ne pas disparaître</a>. This last one, written by Gabriella Papadakis</a>, a former Olympic champion, was very interesting, sad and unfortunately not surprising. It describes how abusive elite level sports can be, and I think it'd be useful for everybody interested in sports to read it. It's so easy to overlook how crazy those folks' lives are, the pressure they face to always perform, be perfect, push through injuries... Her book is one side of the story, and I believe that her former partner is suing her, but what I take from it is that elite sport, however enjoyable to watch, is just insane. Quoting (by memory) Boston Marathon winner Desiree Linden</a> from a recent episode of the Nobody asked us</a> podcast: "Nothing in what we do is normal".</p> I had a good week in terms of running, with 34.61 km, and a very good one at GeoGuessr</a>. I barely played - only 20 games - followed my "at most 30% competitive games" rule, and won 12 out of my 20 games. We'll see next week if it was just a fluke :-)</p> Open source</a> </h1> It was a very light week open source wise, because, again, Olympics :-)</p> In the SourceHut</a> front, I submitted the next iteration</a> of my work to add a GraphQL API to hub.sr.ht, and continued to combat flakiness in sr.ht-integration-testing</a>.</p> I also submitted two fixes to aerc</a>: a regression making the IMAP header cache useless (patch</a>) and a UI freeze when aggressively scrolling through the message list (patch</a>).</p> Status Report - Week 6 2026 Unknown — Mon, 09 Feb 2026 00:00:00 +0000 Real life</a> </h1> We're getting closer to the kids' winter vacation - one more week before we go skiing \o/</p> Speaking of winter, this week was also the start of the Olympics, which already gave me lots of emotions:</p> The French cross country ski team made us proud at the men's skeleton (Hugo Lapalus</a> is 100% my favourite nordic skier), and Klaebo</a> demonstrated once again that he's from another world</li> The French biathlon mixed relay team was outstanding (and Julia Simon</a> is the best ever [1]).</li> </ul> I think that winter Olympics don't make any sense any more, and I am mad that we will organize the 2030 ones in France [2], but since they take place, I'm supporting the athletes with all my heart.</p> I did not only watch TV, and finished Myth of self-esteem</a>, and started Overcoming the rating game</a>. Both might seem old but I find them super relevant.</p> I had an OK week in terms of running, with 24.63 km, and a mediocre one at GeoGuessr</a>, with 6 wins out of my 20 "counting games" (17 out of 36 overall). This was (un)fortunately not bad enough to be demoted to Gold II, which I think would be deserved... maybe next week :-)</p> I feel like I'm talking about a change in my GeoGuessr strategy in every status report, and this week's no exception. I realized that I'm basically never training: all my games are competitive and count for my ELO, which does not make any sense - you have to train to compete! So I'm implementing yet another rule: if more than 2/3 of my games in a day are competitive, my tracking sheet turns all red. We'll see if this "shaming me into training more" strategy works.</p> Open source</a> </h1> This was another quiet week coding wise, with just a bit of work to bring a GQL API to SourceHut</a>'s project hub:</p> I iterated on pre-requisite patches to generalize "Resource IDs" (main patch</a> and follow-up</a>)</li> I sent another iteration of the read APIs patch</a></li> And started to prototype write APIs (pre-requisite patch</a>)</li> </ul> I also spent a bit of time reviewing a bunch of unreviewed aerc</a> patches.</p> --</p> [1] She could be better outside of biathlon; see why she was banned by the French federation from the first 3 world cup stages this year...</p> [2] See for instance Le crépuscule des Jeux</a></p> Status Report - Week 5 2026 Unknown — Tue, 03 Feb 2026 00:00:00 +0000 Real life</a> </h1> It's been yet another pretty boring week, taking the kids to places, getting bikes fixed, fixing meals, etc. I love</strong> boring :-)</p> I spent an awesome weekend in Brussels for FOSDEM</a>, and will talk about it in a dedicated post.</p> I continued reading Myth of self-esteem</a>, that's pretty good, and got me add (so far) 2 more books to my to-read list.</p> I ran a bit less - 22.74 km - courtesy of FOSDEM, but walked more - 12.21 km - so it was an OK week moving-wise.</p> I'm still spending quite some time playing GeoGuessr</a> with mixed results and a lot of frustration :-) I won 6 of my 20 "counting games" (25 out of 56 overall), which keeps me in the Gold I division. I wish I could say I´m stagnating, but I think I'm actually getting worse, so clearly need to do something... I just don't know what :-)</p> Open source</a> </h1> I did not write a lot of code this week, but spent a lot of time on open source via my trip to FOSDEM (again, I'll talk about it in a separate post).</p> On the SourceHut</a> front, I made a bit of a fool of myself here</a>, and spent some time trying to understand why. TL;DR: some code of mine was merged 2+ weeks ago that should not</strong> build, but actually does because it's picked up by the GQL code generator that fixes it on the fly. So no harm done, but I think it sucks to be able to have differences between the checked-in code and what actually runs... So I proposed</a> an improvement to the CI's linting step.</p> In order to test more scenarii around builds.sr.ht in sr.ht-integration-testing</a>, I submitted a patch</a> to hut</a> to support updating build tags.</p> Finally, I spend some time reviewing the (large) aerc</a> patch series</a> to create a local cache for IMAP and JMAP. It's very good and I'm looking forward to it being merged!</p> Status Report - Week 4 2026 Unknown — Wed, 28 Jan 2026 00:00:00 +0000 Real life</a> </h1> This week's been pretty quiet, with a bit more action on the weekend: we had the dog (Fifi) and cat (Chatouille) of friends home. They are very nice, but also like to remind us that they exist at 4 AM... so we were happy to have them, and not sad when they left :-D</p> I have finished reading the excellent Guerres oubliées</a>, and started Myth of self-esteem</a>. The jury is still out about whether I should continue or just stop The Silmarillion</a>; I generally don't like dropping books, but this one is definitely not an easy read!</p> I continue to be pretty consistent in terms of running, with 25.91 km, and thanks to Fifi, add to this 11.93 km of walking.</p> My GeoGuessr</a> week was mediocre, with 7 wins out of 20 "counting games" (19 out of 42 overall), enough to stay in the Gold I division. It's not clear to me whether I'm actually making progress at this game, but I like it so I'll just keep on playing.</p> Open source</a> </h1> My patch</a> to add a GraphQL API to SourceHut</a>'s hub.sr.ht service started a very good discussion about IDs, and Drew introduced</a> the notion of ResourceID. I helped use it in lists</a> and trackers</a>, and submitted a new version</a> of the GraphQL API for hub.sr.ht.</p> I spent a bit of time on aerc</a>, mostly reviewing Robin's big patch series</a> to introduce a client side cache. Testing it highlighted a minor bug in one of the logs I added some time ago, fixed by this patch</a>.</p> Status Report - Week 3 2026 Unknown — Tue, 20 Jan 2026 00:00:00 +0000 Real life</a> </h1> When it's winter, folks are sick and your wife has many students coming to your home for their classes, there's almost no chance you don't get sick as well. Our family followed this rule, with Lyla and Nathalie impacted :'(</p> On the tech side, we bought a new dishwasher 2 months ago since the previous one died and ... drumroll... it's already having problems! The super duper feature to automagically open the door towards the end of a programme to improve drying works so well that we can't close this f.ing door anymore! If you block it with a chair, everything works, so we have a workaround, but we have had to book a repair with Siemens, and someone will come next week. I don't care about fancy features, I just want machines that work - why can't we have super simple ones!?!</p> On the reading side, I have finished the excellently depressing Pauvre petit blanc</a>, and started The Silmarillion</a>. So I have not increased the number of books I'm reading in parallel (remember I'm also</strong> on Guerres oubliées</a>) but I have not decreased either...</p> I had a good week in terms of running, with 7 runs for a total of 37.12 km, and an OK one at GeoGuessr</a>, with 6 wins out of my 20 "counting games" (12 out of 29 overall).</p> Open source</a> </h1> I have finally decided to buy a new laptop, and I have received it this week \o/ This means that I did not contribute much to open source this week, because I spent quite some time to set everything up and tune GNOME</a> to my liking.</p> The good news is that this laptop is vastly more powerful than the pretty old hardware I was working on, so build times for GCC</a> or Servo</a> are good, and I can consider contributing again to those projects!</p> Speaking of GCC, I took a relatively easy ticket</a> as a warm-up exercise, and merged a fix</a>.</p> I still had a bit of time to work on SourceHut</a>, fixing</a> an issue with settings propagation from mailing lists to the projet hub, and submitting an iteration of my patch</a> to bring a GraphQL API to this hub.</p> This last patch triggered a good discussion</a> about IDs and won't be merged, but expect some activity on this topic in the coming days.</p> Status Report - Week 2 2026 Unknown — Thu, 15 Jan 2026 00:00:00 +0000 Real life</a> </h1> The big highlight this week is that we had snow here in the Netherlands, and that it did not melt immediately. To my despair, it had not happened for a few years, so I was really glad :-)</p> </p> The kids were even lucky to have their Wednesday classes cancelled because of the snow! I doubt folks from Norway or Sweden are reading this, but if so, I'm sure they're (rightfully) laughing at us :-)</p> I have finished the first tome of La guerre et la paix</a>, continued reading the excellent Les guerres oubliées</a>, and started Pauvre petit blanc</a>, yet another book I heard about either in Le Collimateur</a> or Paroles d'histoire</a> (as a side note, I probably should focus my reading a bit more instead of starting parallel threads :-))</p> In terms of running, I have not yet broken my goal to run a bit everyday, and ran 28.28 km.</p> Finally, I had ups and down this week at GeoGuessr</a>, with a catastrophic afternoon where I basically lost all my games, and like 100 ELO points... super frustrating. This got me to put in place "gold rules" for myself, in particular to stop playing for at least 30 minutes as soon as I've lost 3 games. It will probably not change anything in the long run (at my super low level, you can as easily earn 100 ELO points), but I'm sure it will be better for my sanity. In terms of statistics, I won 7 of my 20 "counting games, and 22 out of 50 overall.</p> Open source</a> </h1> For once, I did not work exclusively on SourceHut</a>, but also on ActualVim</a>, as described in this post</a>. I also sent a patch</a> to restore the "command line" capabilities with recent (>= 0.11) versions of NeoVim</a>.</p> On the SourceHut front, I submitted:</p> A batch of patches to simplify some well spread overly complicated error handling pattern: 1</a>, 2</a>, 3</a>, 4</a>, 5</a> and 6</a>.</li> A patch</a> allowing users to update the tags of their builds. I will use this to track the failure reason of sr.ht-integration-testing daily test runs.</li> A patch</a> to hut</a> to allow changing one's notification preferences in todo.sr.ht.</li> </ul> Speaking of sr.ht-integration-testing</a>, I added a test</a> for todo.sr.ht´s setting to get notified of your own updates, and did various "maintenance tickets".</p> Having (some) fun with Sublime Text and ActualVim Unknown — Mon, 12 Jan 2026 00:00:00 +0000 I have been using Sublime Text</a> for more than 10 years [1], and I really love it: it's fast, has many extensions, works on all the platforms I use, and my single license allows me to use it on as many computers as I want. If it were open source, it would really be perfect ;-)</p> When I resumed working on GCC</a> 18 months ago, I tried to configure Sublime to support GNU's exotic coding style, and failed. So I reverted to using my old friend vim</a> because the GCC folks ship a vimrc</a> that just does the right thing, but I was a bit frustrated.</p> Fortunately, I stumbled upon ActualVim</a>, a Sublime package that uses neovim</a> as "backend", giving me support for GCC's vimrc within Sublime - perfect!</p> At the time, I was coding 100% on an iMac, and things worked great, and when I resumed (also) using Linux, it worked OK-ish, but not as well, for reasons I did not understand nor really investigate.</p> The two main annoyances were:</p> With /</code> based search: I somehow was not accessing the history via up/down, and backspace was acting weird: for example typing /cou</code> followed by backspace and v</code> would look for couv</code> instead of cov</code>. Annoying but not the end of the world, and I got used to simply escaping the search box and restarting whenever I was typo'ing.</li> With the "Enter" key when editing GCC code: the cursor would always be on the new line's first column, instead of being properly indented. A bit more annoying, but since I've been working on SourceHut</a> and aerc</a> almost exclusively in the last 6 months, it did not really bother me.</li> </ol> Last week, I decided to spend some time to understand what was going on, and why things worked perfectly well on my iMac and not on Linux.</p> A bad patch...</a> </h1> I added a bunch of logs and experimented, but since I was not going anywhere, I decided to just write a patch that would fix my problem, even if imperfect, and send it out for feedback (and also to understand the status of the project, that did not had any commit in the last 8 years).</p> This patch</a> starts from the fact that somehow some keystrokes in Sublime don't make it to neovim, and simply accumulates them in memory, and only send the resulting string to neovim when the user validates the search.</p> Since it worked for me, and I assumed that I might not be the only one in the world seeing this issue, I sent the patch, ready to never hear back - again, there's not been any change in 8 years, and as much as I love it, Sublime is not the most used code editor in the world...</p> To my surprise, I got a reply from the maintainer within a few hours! We discussed a bit in the pull request, and agreed that it was problematic, so I discarded it.</p> What I find awesome here is first that the maintainer spent time to look at a random pull request in a project he has probably not looked at for years, and second that something in the discussion (not sure what) got me to think a bit more and fully understand the problem.</p> ... to unblock everything</a> </h1> Sublime allows you to customize keyboard shortcuts, and I realized that I was using this on my Linux installations, but not on MacOS - because the defaults for Linux did not match the MacOS ones, that I'd gotten used to, hence needed changing. Here was the difference, finally!</p> I dug a bit deeper in those customizations, particularly the priority if the mapping for a given key sequence was different in the default keymap, the package's keymap and the user override.</p> This post</a> explained this, and gave me a nice workaround. But first, what was going on?</p> Since I needed to customize some key bindings, my user keymap was a modified copy of the default one, with an action defined for the backspace key (as well as up, down and Enter). Per the priority rule, this is the mapping that was taking precedence, not the one defined by ActualVim, that says to forward the key sequence to neovim.</p> So whenever I typed backspace, up, down or Enter, the event would be handled natively by Sublime, and not sent to neovim, leading the exactly what I was seeing!</p> Knowing this, the fix is easy: create a symlink to ActualVim's keymap in my user directory, so that it takes precedence over the customized defaults. Et voilà, mystery solved!</p> Key takeaways</a> </h1> This ended up being a fun adventure, and here are the main things I take away:</p> Don´t assume that you won't get help from maintainers of seemingly dead projects; just ask questions or start a discussion, ideally with a patch, even if imperfect: worst case you don't hear back, but chances are that you do get help.</li> The urge to investigate when things are weird can cost (a lot of) time, but your world is better if you get to the bottom of things and restore balance in the force :-)</li> </ul> --</p> [1] According to my archives, I bought my first license in February ... 2014.</p> Status Report - Week 1 2026 Unknown — Tue, 06 Jan 2026 00:00:00 +0000 Real life</a> </h1> We took the plane back to the Netherlands on January 1st, a very good idea since it's been snowing here since then, seriously impacting flights and trains. It's been years since we did not have snow, and I am very happy and hope that it will not only last for a few days, but rather that we have a "real winter" for once.</p> I continued reading La Guerre et la paix</a>, and started Les guerres oubliées</a> in parallel, by my two favourite historians, Alexandre Jubelin</a> and André Loez</a>.</p> I'm slowly getting back to running, with 18.12 km this week, and an objective to run more than 6 days a week (in average) this year.</p> I had a good week at GeoGuessr</a>, with 12 wins out of my 20 "counting games" (22 out of 37 overall), which gets me back into the Gold I division \o/</p> Open source</a> </h1> I submitted a few patches to SourceHut</a>:</p> Fixed</a> a regression that I introduced just before Christmas with multiline selections in git.sr.ht</a> (I did not catch it in my manual testing because of a cached APK in my sr.ht-container-compose</a> setup :-/). This got me to add tests</a> to cover those cases in sr-ht.integration-testing</a>.</li> Improved todo.sr.ht</a>'s GraphQL API to expose</a> user preferences.</li> </ul> I also communicated the December release</a> of sr.ht-integration-testing</a>, and made a few good improvements:</p> Added basic tests</a> for paste.sr.ht</a>.</li> Improved the infrastructure</a> to allow asserting that outbound emails are actually sent.</li> </ul> Status Report - Week 52 2025 Unknown — Wed, 31 Dec 2025 00:00:00 +0000 Real life</a> </h1> We flew to Villefranche-sur-Saône on Wednesday, and have been staying at my mom's there since. Our main activities have been eating, drinking, discussing and sleeping.</p> Apart from that, I continue my slow progress through La Guerre et la paix</a>, and finished Game Engine Black Book: Doom</a>. This book was really excellent, and I think I will read the author's book about Wolfenstein 3D next.</p> I fought hard not to register a 0 km running week, and ran 5.14 km. And what was doomed to happen for a few weeks did happen: I was demoted to Gold II at GeoGuessr</a>, with only 4 wins in my 20 "counting games" (23 out of 52 games overall). I´m a bit disappointed since I've been in Gold I for around 2 months, but I've been saying that folks there are better than me for a few weeks, so it's only logical.</p> Open source</a> </h1> This was a real vacation week: I did no work whatsoever \o/</p> Status Report - Week 51 2025 Unknown — Tue, 23 Dec 2025 00:00:00 +0000 Real life</a> </h1> After yet another trip to Lille on Monday, this week was a quiet winding down to our 2 week Christmas vacation. We'll spend most of it with family in France.</p> I am making steady progress on La Guerre et la paix</a>, that is consistent with my limited experience with russian literature: it's rather long, very slow - and barely anything happens - but it´s still super enjoyable to read. As I was telling Nathalie: you have to force yourself a bit to pick up the book, but then it's hard to put it down - weird :-)</p> I'm interleaving my reading with a very different and also excellent book: Game Engine Black Book: Doom</a>. It's not only about the engine itself but also the state of hardware circa 1993; I really like reading about this "older stuff".</p> 2025 will have surely not been the year of running: I only ran 17.76km this week, and did not particularly enjoy it. I hope 2026 reignites my "running is stupid but I love it" fire.</p> Finally, I had a good week at GeoGuessr</a>, staying in the Gold I division after winning 7 of my 20 "counting games" (and 42 out of 91 overall). The numbers don't show it yet, but I do feel that I'm making progress. Time will tell.</p> Open source</a> </h1> As I mentioned in this post</a>, I have migrated my old iMac to Fedora, and I continue to be very happy. It allowed me to do all that follows without having to touch Anton's PC - it's easier for me, and I'm sure he's grateful to not need to kick his dad out of his room anymore.</p> On the SourceHut</a> front, I did a bunch of random things as usual:</p> I submitted a patch</a> to surface the "superseding patch" in the patchset view when applicable. It makes my life easier because I tend to send multiple patch revisions, and I hate having to search to find the latest one. The UX is pretty nice - this is the blue notice box (fortunately Drew had a better idea than my initial one :-))</li> </ul> </p> I investigated a weird bug with build badges when several are included in the same README, that ended up due to a bug in the SVG cache key we use (fix</a>).</li> I submitted a fix</a> to hub.sr.ht so that it properly sets up the webhooks it registers to itself in builds.sr.ht.</li> I removed</a> linelight.js from git.sr.ht since it's now</a> shipped with core.sr.ht.</li> For once I read some documentation :-) and fixed a typo</a>.</li> </ul> Thanks to the hub.sr.ht bugfix mentioned above, I'm finally convinced that sr.ht-container-compose</a> does not need this patch</a>. This means that I can reduce the number of changes between my fork and main \o/ Once</a> all</a> those</a> patches</a> are merged, the only differences will be in the Makefile.</p> And I made a few changes to sr.ht-integration-testing</a>:</p> Significantly speedup (and robustifed) builds related tests via this ticket</a>.</li> Added a test</a> for my recent fix about empty mbox downloads.</li> I retried running the testsuite natively (i.e. on SourceHut builders) instead of on EC2, and it still does not work</a>. At least I have the reason documented in a ticket, and don't need to try again in a few weeks to relearn it :-)</li> </ul> I took a look at one Servo ticket</a> that might explain why sr.ht</a> does not render well with Servo</a>, but did not make any progress beyond pulling some logs.</p> Giving a second youth to an old iMac that did not want to die Unknown — Fri, 19 Dec 2025 00:00:00 +0000 I have a bunch of computers at home, and the one I am primarily working from is an old 27-inch iMac from 2013 (!) [1]</p> </p> It's not super fast, but I really like it, and it was alright for all my development tasks (GCC</a>, aerc</a>...) until I started to need Docker (for SourceHut</a> testing), that I never managed to install on Catalina (the last MacOS version that Apple allowed me to install).</p> It's really nice to have hardware run great [2] for so long in this world of planned obsolescence</a>, and while not blocked in anyway (I do have more recent computers :-)), I was frustrated not to be able to use this machine fully.</p> I've been reading about folks installing Linux on Apple hardware, but I was afraid to brick everything, so constantly postponed trying. This week I finally took the plunge and decided to install Fedora</a>...</p> Installation adventures</a> </h1> I met the first hurdle after I downloaded the "Fedora Workstation" ISO</a>: finding a USB stick to write it to...</p> A dog to the rescue</a> </h2> I'm pretty sure I have many USB sticks in the house, but could not put my hands on any, so I decided to go buy one at the local Action</a> store, and couple this with a short run. Surprinsingly enough, Action was out of USB sticks (!!), so I tried Hema</a> and only found a ... dog USB stick</a>. It's not nice, takes up two slots because of its "big" bottom, but it does the job...</p> </p> Where we're going, we don't need roads^W cables [3]</a> </h2> I have a bluetooth Apple keyboard and a Magic Mouse, which is a problem when you setup a brand new computer, so I had to temporarily steal Anton's keyboard and mouse, plug them along with the dog, and boot things up.</p> Still afraid to brick everything, I started in "live mode" to look around, and see that I could at least detect and use my bluetooth keyboard and mouse, before actually erasing my disk and installing the OS. It worked surprisingly well, which gave me confidence to go ahead.</p> The installation went smoothly and I was happy, until I tried to connect to the wifi. It turns out that the iMac uses a Broadcom wifi adapter, that requires a proprietary driver, definitely not</strong> included in the ISO.</p> So I had to take the iMac downstairs and plug it directly to the Ziggo box to get the wifi working.</p> I have no tools because I destroyed my tools with my tools [4]</a> </h2> Fortunately Apple does not use some weird Ethernet adapter, and I could easily find out how to download and setup the right kernel modules here</a>. It went well, but since I'm very good at skimming through documentation instead of actually reading it, I made two mistakes:</p> I did not think that sudo akmods</code> failing might be a problem (in hindsight, lol)</li> Even though I had pulled lots of updates (including kernel ones), I thought that rebooting was a good idea</li> </ul> When the machine went back up (with a new kernel), I did not have bluetooth anymore (and still no wifi).</p> After replugging Anton's keyboard and mouse, learning how to change the booted kernel version, a reboot, installing kernel sources and this time properly building kernel modules, I was back to business, with bluetooth and wifi working!</p> I bid farewell to wired accessories, took the iMac back upstairs, and started setting things up.</p> Configuration adventures</a> </h1> Overall, getting up to speed was much faster than I was expecting, so I will only talk about two things.</p> Adapting to muscle memory</a> </h2> The human body and mind are incredible things: even though I am working from at least 4 different computers at home (two iMac-s, one Windows PC, and one Windows laptop, each with their own specific keyboard layout), my fingers automagically adapt to whatever is in front of me.</p> It was therefore important for me to be able to not use "CTRL+" shortcuts but their Mac equivalent "Command+".</p> Thanks to Gnome Tweaks</a>, it's actually very easy: all it takes is to use the "Ctrl is mapped to Win and the usual Ctrl" option.</p> And by overriding the Command+Space shortcut to "Search" in Gnome's regular "Keyboard shortcuts" preferences, I almost feel like I'm still on Mac OS.</p> My life was about upgrading, now it's about downgrading</a> </h2> The second noticeable configuration adventure is a fun one, and marks a big shift for me: after years trying to upgrade software by hand (because I was on an outdated MacOS version), I now need to downgrade some software by hand :-)</p> For context, I'm a big Sublime Text</a> fan, and in order to have it properly format code for GCC development</a>, I use the ActualVim</a> package. ActualVim delegates all editing actions to an instance of NeoVim</a>, and makes Sublime do exactly what I need <3</p> However, on the "new" iMac, things were only partially working; in particular, typing /</code> would not do anything, instead of opening a "search box". A bit of debugging showed that ActualVim was expecting 6 elements in the cmdline_show</code> UI event, and was getting 7 from NeoVim. It turns out that Fedore 43 ships Neovim 0.11, that has a related API change</a>!</p> Since ActualVim is pretty old and likely not maintained anymore, the easiest was for me to build NeoVim 0.10 from sources.</p> I have no idea whether there are 3 of us in the world using ActualVim or thousands (in other words, whether it's worth fixing), but I plan to investigate and if it makes sense, to try to fix this.</p> Thoughts about the move</a> </h1> I'm writing this on my third day after installing, and I am very, very happy to have done it: things are working surprisingly well, and I am able to do all</strong> the coding and testing I want directly on the iMac.</p> I still have a few minor things to sort out (e.g. be able to change screen brightness from the UI and not by writing numbers in obscure files in /proc</code> :-)), but nothing important nor urgent.</p> The old iMac is dead, long live the old iMac!</p> --</p> [1] We have a more recent one, but it's used primarily by Nathalie for her work so I don't want to mess with it :-)</p> [2] I did have to change the hard drive a while ago, but that's the only issue I got in all these years.</p> [3] Quote from the "Back to the future" movie.</p> [4] Quote from James Mickens' hilarious The Night Watch</a>.</p> Status Report - Week 50 2025 Unknown — Tue, 16 Dec 2025 00:00:00 +0000 Real life</a> </h1> Everybody at home is looking forward to the Christmas break; two more weeks!</p> I've gone to Lille for the second time in 2 weeks, and have another trip there scheduled next Monday; we're probably at a point where the car could do the driving on its own :-) It takes a lot of time and is a bit tiring, so less work for me, but it's alright - let's say I´m warming up for the vacation.</p> I finished reading Emmanuel Carrère's latest book, Kolkhoze</a>, that was excellent as usual. From my point of view, Carrère is one of the greatest authors alive. I also read Task Force Baguette</a>, after one of its authors was invited to Le Collimateur</a>, and it was OK. And since Carrère was talking about how his uncle became a big fan of Leon Tolstoï, I decided to start reading La Guerre et la paix</a>, that should keep me occupied for a bit :-)</p> It was another mediocre week in terms of running (20.49km run), and a good one at GeoGuessr</a>: despite winning only 5 of my 20 "counting games" (29 out of 59 overall), I stay in the Gold I division, and I reached 701 ELO. I expect to drop a bit next week, but am happy to have reached a personal best.</p> Open source</a> </h1> I continued to make miscellaneous improvements to SourceHut</a>, in particular:</p> Support selecting multiple lines in pastes.sr.ht, via this</a> and that</a></li> Expose service version and build date via GraphQL in builds</a>, git</a>, hub</a>, lists</a>, paste</a>, todo</a>, man</a> and hg</a></li> </ul> I also prototyped adding read-only GraphQL APIs to hub.sr.ht, via (work in progress) patch</a>.</p> Along the way, I sent one patch</a> to hut</a> to support more services in hut graphql</code>, and the associated configuration update</a> in sr.ht-container-compose</a>.</p> Finally, beyond miscellaneous small fixes to sr.ht-integration-testing</a>, I finally split</a> the pure build and the "testing SourceHut via sr.ht-container-compose". I noticed a weird bug with build badges, that I sent a fix</a> for.</p> Status Report - Week 49 2025 Unknown — Tue, 09 Dec 2025 00:00:00 +0000 Real life</a> </h1> The main event this week is that we are now the owners of the apartment we've been renting for a few years - we were not particularly looking to buy, but circumstances were such that it was the right thing to do, and we're happy.</p> It was also the opportunity to discover a whole ecosystem of people you have to give money to to be able to buy, and whose "added value" is not always very clear... a topic for a separate post :-)</p> This weekend, we welcomed the pets of friends, Fifi (dog) and Chatouille (cat), and it was a great way of showing the kids that having pets is quite some work... I have good hopes that the "Why can't we have a dog?" discussions stop for a little while.</p> I finished reading La Révolution française</a>, that I did not like very much; I think I tried to punch above my weight: my knowledge of the period is too limited for this book to be "useful" to me.</p> I also read Et tuez-les partout où vous les trouverez</a> by Abou Djaffar (who I discovered in an episode of Le Collimateur</a>), that was excellent.</p> I did not run much (18.47km), but walked a lot more than usual (16.12km), courtesy of Fifi :-)</p> Finally, my GeoGuessr</a> week was good: I finished 18th at my Gold I division, which is mediocre but I'm happy enough with. And I enjoy my new strategy to not worry too much about the 20 first games (that I generally lose because my opponents are much better than I am) and play a lot more (against people at my level). I'm a lot less frustrated, and feel that I progressing more - win-win!</p> Open source</a> </h1> It was a quiet week in terms of coding, partly because of housing stuff, pet stuff, watching some re:Invent keynotes</a>, and fighting with disk space (yeah I'm looking at you stupid WSL ticket</a>).</p> Still I did a few things, in SourceHut</a> as usual:</p> Sent (I hope) final revisions of changes to expose the running version and its build date in GraphQL (this</a>, this</a> and that</a>).</li> Fixed spurious errors</a> during hut export</code>export.</li> Made minor</a> fixes</a> to hut</a>.</li> </ul> Status Report - Week 48 2025 Unknown — Mon, 01 Dec 2025 00:00:00 +0000 Real life</a> </h1> The highlight of the week was the Gojira</a> concert in Lille on Friday evening. While it was my third (or fourth?) time seeing them (the first being 20 years and 3 days before), it was the first for Lyla and Anton, who really liked it. The band was excellent, as usual, and I'm looking forward to (at least) 20 more years regularly attending their concerts :-)</p> Here are a few pictures. Note the drone powered whale on "Flying whale" :-)</p> </p> </p> </p> I finished reading Par le fer et par le feu</a>, and learnt a lot about naval combat on the Atlantic ocean in the 17th and 18th centuries (sic). I continue in the field of history, with La Révolution française</a>, that might confirm my suspicion that the situation in France and the world is becoming close to what it was a bit before 1789... we'll see.</p> My running continues to be mediocre, with a mere 13.4 km this week. It's partly explained by my having more fun coding than running, and I need to balance things out better.</p> Finally, my GeoGuessr</a> week was not great, but OK. I have realized that I don't really deserve to be in the Gold I division (in other words, the folks there are much better than I currently am :-)), so my strategy is now to:</p> Quickly get it over with the first 20 games, that do count for the week's division ranking, and are against folks from that division,</li> Then play a lot more games, that don't count for the division ranking (but still do for my ELO) and are more useful and interesting, because I'm playing against folks with a level close to mine.</li> </ul> Open source</a> </h1> As usual, the bulk of my work was in the SourceHut</a> ecosystem:</p> The work to allow users to control what actions trigger builds is merged \o/ The main patches are git.sr.ht</a>, hub.sr.ht</a>, and sr.ht-docs</a></li> Right after, I migrated</a> sr.ht-integration-testing</a> to use this instead of the hacks I had written</li> I resurrected</a> my work to expose build information (e.g. git hash and build date) in the version</code> GraphQL API.</li> </ul> I also announced</a> the first monthly release of sr.ht-integration-testing</a>, and took the opportunity of my annual plan with IRCCloud</a> expiring soon to migrate everything IRC to chat.sr.ht</a> (and that's another 5-ish bucks saved per month :-)).</p> Status Report - Week 47 2025 Unknown — Mon, 24 Nov 2025 00:00:00 +0000 Real life</a> </h1> This was yet another quiet week on the personal front - this is good.</p> I have finally finished reading The Lord of the Rings</a>, and really liked it. I completely ignored this book's existence for decades because my contrarian self despised all the fuss people were making about the movie series 20+ (!!) years ago, and it was a mistake. Its size (1000+ pages) can be scary, but its does not "feel" long, and the story is really good. Thumbs up.</p> I have started Par le fer et par le feu</a> by Alexandre Jubelin, of Le Collimateur</a> "fame", initially as a means of supporting his work, but it turns out that the history of naval combat in the 16th and 17th century is actually super interesting. I'm learning a ton of "not super useful", and loving it!</p> I ran 23.6km, enjoying the colder temperatures. While I'm still pondering whether I should sign up for the Leiden Marathon this year, Lyla and I have for the 15km race at the CPC Loop Den Haag in March.</p> My GeoGuessr</a> week was painful, even though I remain in the Gold I division: I lost 36 of my 63 games, and felt desperate during a few "catastrophic loss streaks". There's a saying in French that I don´t really know how to translate describing the week well: "c'est le métier qui rentre".</p> Open source</a> </h1> I continued to work mostly on SourceHut</a> and aerc</a>, but I am starting to feel a itch to do some GCC</a> development again; time will tell if it's just talk or it actually happens.</p> On the SourceHut from, I got a cool shout-out in their Q4 What's cooking on SourceHut?</a> post, which is very nice.</p> I iterated on the "submitter patch series", that give users more control on what events trigger builds. For instance, you'll be able to state that you only want builds to happen for pushes to master, and not for patchsets received on a mailing list. This</a> and that</a> patch are merged, and I'll send new revisions for this</a> and that</a> today.</p> I also sent a trivial compilation fix</a> identified by sr.ht-integration-testing</a>.</p> Talking about sr.ht-integration-testing, I implemented a bunch of semi-random improvements, mostly to simplify things and combat flakiness:</p> Use hut instead of the UI when possible for builds.sr.ht (ticket</a>)</li> Tighter output for daily runs (ticket</a>)</li> Support using my forks instead of upstream in sr.ht-container-compose</a> if needed (ticket</a>)</li> Use require</code> instead of assert</code> in tests (ticket</a>)</li> Move git related functions into a class (ticket</a>)</li> </ul> Those require a small improvement to hut</a>'s hut builds list</code> (patch</a>). Along the way, I noticed that hut builds resubmit</code> did not conserve build tags, just like the UI did, hence this patch</a>.</p> Finally, I spent a bit of time on aerc, in particular investigating an issue recently reported on IRC, and again in the mailing list</a> with one of the features I implemented way back. The reporter did an awesome detective work, and I was mostly the scribe in the fix</a>.</p> In passing, I added an option to get go-imap</a> to log its interations with the IMAP server (patch</a>) and made a pretty useless error log a bit better (patch</a>).</p> Status Report - Week 46 2025 Unknown — Tue, 18 Nov 2025 00:00:00 +0000 Real life</a> </h1> This week I spent quite some time at my kids' school: each had their "parent/teacher" meeting (that went unsurprisingly very well), and I participated to an "interview panel training" for last year high schoolers on Saturday.</p> I have a lot of opportunity to witness the sad state of education via Nathalie's business, and this session demonstrated two things yet again:</p> The number of no-shows tells me that many students have 0 sense of responsibility, and have no problem cancelling engagements at the last minute because they just don´t feel like going...</li> I'm 100x more hopeful for girls to succeed in life than boys. Boys are really lucky that society managed to shatter girls' self-confidence; when we manage to fix this, girls will simply destroy them.</li> </ul> I made good progress at The Lord of the Rings</a>, and am pretty hopeful I will finish it this coming week. It will be good because the pile of other books I want to read is quickly growing ;-)</p> As usual, I listened to many podcast episodes, but the highlight this week was the discussion Steve Huynh had with Philip Su in his A life engineered</a> podcast (you can watch it on Philip's Substack</a>). I really love listening to Philip, and things clicked particularly well with Steve, giving us an oustanding episode.</p> I did not run that much this week - 19.88km - and had pretty good one at GeoGuessr</a>, finishing the week ranked #9 in my Gold I division.</p> Open source</a> </h1> I improved sr.ht-integration-testing</a>'s daily test runner</a> so that it's only triggered for pushes to master</code> (I use branches quite extensively, and got sick of adding -o skip-ci</code> to most my git push</code> :-))</p> Funnily enough, I then discovered this SourceHut ticket</a> that aims at doing something very similar, so took a stab at it. We're still iterating on the code review</a>.</p> Still in the SourceHut ecosystem, I fixed a bug</a> in my recent work on this ticket</a>, fixed another schema.sql synchronization issue</a> and send two patches to hut</a>: this</a> and that</a>.</p> Status Report - Week 45 2025 Unknown — Sun, 09 Nov 2025 00:00:00 +0000 Real life</a> </h1> We're back to our normal routine after the very nice break in France, and the week was uneventful.</p> I continue reading The Lord of the Rings</a>, and am almost at the end of "The Two Towers" - with a bit of luck I'll be done by the official end of the week :-)</p> I am happy to be back to more serious running, and that none of my knees caused any problems in spite of going from 0 to 25.94km this week.</p> In terms of GeoGuessr</a>, it was very good: I was glad to see on Monday morning that I had been promoted to Gold I, and I had a pretty good week, with 8 wins out of 20 duels. It's definitely not elite level, but after my recent "19 losses out of 20" truama, I'm satisfied. I will only know for sure on Monday morning, but I have good hopes to remain in Gold I next week - fingers crossed!</p> Open source</a> </h1> I resumed working in the SourceHut</a> ecosystem, in particular my sr.ht-integration-testing</a> package.</p> I spent some time to create a daily job to run the testsuite with SourceHut master and sr.ht-container-compose</a>, with the hope to spot possible regressions within 24 hours, whether I'm active or not. With the help of Lumo</a>, I setup the following: every day at 08:42 UTC, Amazon EventBridge</a> invokes an AWS Lambda</a> that submits a build of sr.ht-integration-testing with hut</a>.</p> After fixing a few mistakes in what Lumo was telling me to do, I'm happy to say that things seem to work pretty well. If you're interested, you can find the build results here</a> and a description of the setup here</a>.</p> Along the way, I submitted a fix for yet another SourceHut regression here</a>, wrote a patch</a> to ensure that resubmitted builds conserve the tags of the initial build, and sent a follow-up patch</a> to allow users to configure whether they want to receive a copy of their own mails to a SourceHut mailing list.</p> Finally, I looked at the code of a GeoGuessr related website I recently discovered, GeoLeaderboard</a>, because it's super useful but also pretty slow. I submitted a patch</a> to improve the search box' responsiveness, and prototyped</a> a 3.4x reduction of the initial load time of the leaderboard.</p> I sent an email to the website's author to "propose my services" in productionizing the prototype and also improving the reliability of the (I suspect currently manual) daily updates. We'll see where this goes...</p> Status Report - Week 44 2025 Unknown — Mon, 03 Nov 2025 00:00:00 +0000 Real life</a> </h1> We continued our camping trip, and went to Les Calanques de Cassis</a>, then Marseille before visiting my in-laws in Montpellier and going back home. Here are a few pictures.</p> From our hike in Les Calanques, from Luminy:</p> </p> </p> </p> </p> </p> From our visit to Marseille, in particular Notre-Dame de la Garde</a></p> </p> </p> </p> In terms of reading, I'm making steady progress in The Lord of the Rings</a>.</p> Along with our 9.91km of hiking, I went for a short (4.23km) run, and my knee did not complain \o/</p> And finally, I had a good week in GeoGuessr</a>, winning more than half my duels, which got me promoted to the Gold I division.</p> Open source</a> </h1> I did a lot more than last week, but still not that much.</p> Thanks to sr.ht-integration-testing</a>, I identified two recent regressions in SourceHut</a>, for which I submitted this</a> and that</a> fix.</p> One of my actions for the upcoming week is to setup a daily run, so that we catch such things right after they're merged, not only when I manually run the testsuite.</p> I also addressed an aerc</a> issue I recently noticed when changing labels: there should be feedback in the UI, but there was none... The fix</a> turned out to be super simple.</p> Still in aerc, I investigated and fixed</a> a missing notification issue recently mentioned on IRC.</p> Status Report - Week 43 2025 Unknown — Mon, 03 Nov 2025 00:00:00 +0000 Real life</a> </h1> The kids and Nathalie are on vacation for 2 weeks, and we decided to go for yet another camping trip in France, towards Les gorges du Verdon</a> and hopefully higher temperatures.</p> Even though our beloved old (16 years old!) car decided that it was a good idea to intermittently switch on its light at night - and empty its battery - we had a great vacation.</p> In this first week, we camped in La Palud sur Verdon</a>, that sits at 900m of elevation (hence chilly nights in the tents :-)), after stops in Troyes</a> and at my mum's in Villefranche sur Saône</a>. Below are a few pictures from our hikes.</p> First day, from Le chalet de la Maline:</p> </p> Second day, from Le point sublime</a>:</p> </p> </p> </p> I continued to read The Lord of the Rings</a>, had an impressive 0km of running, and went back to the Gold II GeoGuessr</a> league, after my humiliating 19 losses out of 20 duels last week...</p> Open source</a> </h1> No coding at all that week (!)</p> Status Report - Week 42 2025 Unknown — Mon, 20 Oct 2025 00:00:00 +0000 Real life</a> </h1> In the last week before the kids' holidays, I struggled with a weird knee pain, and helped friends who're moving empty their apartment (yeah, it's not super compatible...).</p> I've been waking up earlier and earlier in the last few weeks, and I even "did Pi" - 3:14 - this week :-/ Fortunately, the upcoming time change will help.</p> I continue to enjoy (a lot) reading The Lord of the Rings</a>. I'm only one third through, so this will keep me busy for a few more weeks...</p> Because of my knee pain, I only ran 5.80km; this should in theory have given me more time to study and play GeoGuessr, but it did not... And after 11 weeks in the Gold II division, I had my worst week ever, with 19 (!!) losses out of the 20 ranked games I played. I have no idea what happened, and to be honest, it's affecting me more than it should... I hope it's the kick in the butt that I needed to actually work at my game.</p> Open source</a> </h1> My focus this week continued to be in the SourceHut</a> ecosystem.</p> I worked on a behaviour of lists.sr.ht</a> that's been bothering me for a while: if you send an email to a list, you will not get a copy, even if you're subscribed to that list. Enters this ticket</a>, for which I submitted this patch</a>. And it obviously has a companion test</a> in sr.ht-integration-testing</a> :-)</p> Talking about sr.ht-integration-testing, I spent a lot of time trying to get it to build itself and run all the tests against a sr.ht-container-compose</a> instance upon push... and failed. I think that there's too many layers of virtualization involved, and I decided to setup a VM on EC2, and leverage it from builds.sr.ht</a>.</p> It will probably cost me 2 dollars per month, money well spent to have the integration testing framework automatically test itself :-) and send reports to that mailing list</a>.</p> The build file can be found here</a>, and the scripts it uses there</a>.</p> Status Report - Week 41 2025 Unknown — Tue, 14 Oct 2025 00:00:00 +0000 Real life</a> </h1> We're making good progress on a project that's been keeping us busy lately, and that I'll talk about in a few weeks (I don't want to jinx anything ;-)).</p> I speed read through Personal</a>, that was very good, to start The Lord of the Rings</a>. I have never read it, and did not particularly like the movies, but listening to an episode</a> of Le collimateur</a> about it motivated me to give it a try. I have to say that I love it so far...</p> It was a quiet week in terms of running (19.15km only) because I did a lot of open source work (see below).</p> I was once again disappointed to fail to promote to Gold I at GeoGuessr</a> :'( I'll keep trying...</p> Open source</a> </h1> Three weeks ago, I started to build an automated testing framework for SourceHut</a> because as motivated as I am to help make it even better than it is, I got tired of manually testing my changes, and figured that I'm probably not the only one :-)</p> After a lot of fighthing with Docker on my exotic setups (I code either on Mac OS or on Linux through Windows Subsystem for Linux, and test with Docker Compose through sr.ht-container-compose</a>) and with making chromedp</a>-based tests stable, I was proud to announce</a> sr.ht-integration-tests</a>, and happy to hear from SourceHut's creator that it's a good idea :-)</p> I'll do what I can to make this project successful, but even if nobody but me uses it, it's making my "SourceHut contributor life" much easier; already a good win.</p> On top of getting me to fix more bugs (e.g. this</a>), it allowed me to write tests for a pretty big change around event deduplication</a>, and improve sr.ht-container-compose to support more complex service integration. I would not have been able to root cause and fix</a> this bug</a> without sr.ht-integration-testing.</p> Finally, I also submitted a fix</a> for an aerc</a> crash that could happen when loading a new message while also trying to display an arbitrary header of its.</p> Status Report - Week 40 2025 Unknown — Wed, 08 Oct 2025 00:00:00 +0000 Real life</a> </h1> Things have been pretty quiet, with the normal share of administrative things to take care of, and quite a bit of helping Nathalie handle the success of her maths tutoring business.</p> The kids will be on vacation in a couple of weeks, and I'm looking forward to our upcoming camping trip (that I still need to organize...).</p> I've started reading yet another Jack Reacher novel, Personal</a>, that I find better than the last 2-3 I read. We watched No country for old men</a> with the kids, and it was even better than I remembered.</p> In terms of running, it was a normal week, with 24.15km, and I was hopeful to finally climb to Gold I at GeoGuessr</a>, but I fell short unfortunately... I'll keep trying :-)</p> Open source</a> </h1> I continued to work primarily on my upcoming proposal to add integration testing to SourceHut</a>, in particular trying to work around the fact that not all services provide a GraphQL API. This should be fixed eventually, but I think we need tests now</strong>, not in 6 months, so I bit the bullet of interacting with such services directly through their website using chromedp</a>. After many WTFs, it's now pretty stable, and I'm not far from being ready to publicize this work.</p> Along the way, I submitted several "consistency improvement" patches to hut</a>: patch</a>, patch</a>, patch</a> and patch</a>, as well as improvements to sr.ht-container-compose: adapt to recent master changes</a> and improved logging of Python services</a></p> Another advantage of this effort is the ability to easily validate that the various SourceHut services' master</code> branch interact well with others. On Saturday, things started to break (in the end it was a local configuration issue), and I realized that there's no way of knowing what version is running in Production. I propotyped</a> a GQL API to address this; since it did not get pushback, I'll try to finalize it next week.</p> Finally, I submitted a fix to aerc</a>'s folder name completion when accents are involved (yes, I'm French, I like accents ;-)): patch</a>.</p> Status Report - Week 39 2025 Unknown — Sun, 28 Sep 2025 00:00:00 +0000 Real life</a> </h1> We're making good progress on a personal project I'll talk about when it (hopefully) becomes reality, and induces a bit of stress but for "a good cause". Temperatures are slowing decreasing, and it makes me very happy - I should soon finally be able to run without profusely sweating :-)</p> I did not read much because of the aforementioned project and a good number of bugs that needed fixing (see "Open source" section"), but still made a dent into L'homme sans nom</a> that I should finish next week.</p> Apart from that, it was an OK week in terms of running, with 25.95km, and not too bad at GeoGuessr</a>, though I seem to be stuck in the Gold II division. It's not super surprising however since I "just play", and don't do any studying to actually improve...</p> Open source</a> </h1> This week I started working on a proposal to add integration tests to SourceHut</a>, mainly because I am getting tired of manually testing my patches, but also because there have been a few regressions lately that would have been avoided if we had an easy to run set of basic tests.</p> This effort is nowhere near ready, and I've not even discussed any of this with the SourceHut folks, but I think it's fun (who knows, if I'm successful, I might do the same for aerc</a> :-)) and will be useful at the very least to me.</p> This also got me to play a bit more with hut</a>, and submit a few changes:</p> Don't only rely on stdin for some commands (patch</a>)</li> Expand ~</code> in --config</code> values (patch</a>)</li> </ul> and identify and fix a few of the previously mentioned regressions:</p> Reinstate building sr.ht from scratch (patch</a>)</li> Reinstate wiki (patch</a>)</li> Reinstate updating tickets via email (patch</a>)</li> </ul> as well as an issue in sr.ht-container-compose</a> following recent SourceHut changes (patch</a>).</p> I also submitted the next iteration of my work to deduplicate events in the Activity page (patch</a>), which closes the SourceHut part of this update :-)</p> I worked a bit on aerc</a>, namely</p> Fixed a possible crash handling mouse events (patch</a>),</li> Fixed message opening by clicking on it at the first load of a directory (patch</a>)</li> Fixed a crash in the generic list box, reported on the mailing list (patch</a>)</li> </ul> Status Report - Week 38 2025 Unknown — Tue, 23 Sep 2025 00:00:00 +0000 Real life</a> </h1> This was yet another pretty uneventful week, except for the small heartbreak because the French ladies rugby team lost to England in the World Cup semi-finals on Saturday. It was a good game, and I'm sad to say that the better team won. The squad has a nice game ahead of them against New Zealand to decide who finishes third in the world.</p> I finished reading Make me</a>, yet another Jack Reacher novel. I must say that at the 18th or so in the series, things get a bit repetitive, but still entertaining... we'll see how #19 is :-) And I finally finished How propaganda works</a>, not excellent but still interesting enough for me to not give up.</p> In terms of running, it was an average week, with 22.36 km, and for GeoGuessr</a>, I ranked #9 in my Gold II pool; I was one place away from promoting to Gold I, so I'm hopeful it will happen soon.</p> Open source</a> </h1> I continue to spend a lot of my time addressing issues here and there in SourceHut</a>, and this week was a lot about the projects</a> part:</p> I iterated</a> on the events deduplication ticket</a>; the initial data model is not great IMO, and there's a lot of existing data so coming up with something nice and not too complex is a bit of a challenge...</li> Along the way, I noticed a few minor bugs that I fixed: Webhooks were not properly removed from the database upon project deletion (patch</a>)</li> Activity pages were not consistent in the UI; now they are :-) (patch</a>)</li> sr.ht-container-compose</a> was not properly configuring hut</a> for all the services (patch</a>)</li> hut</a> commands were not 100% consistent with to set an entity's description (patch</a>)</li> </ul> </li> </ul> On Sunday, I decided to team up with Lumo</a> to automate building a heatmap for my open-source contributions across projects / forges. The result is not bad: see the "Heatmap" folded section here</a>.</p> I don't like magic... Unknown — Thu, 18 Sep 2025 00:00:00 +0000 In the last few weeks, I spent a good part of my "work time" trying to improve SourceHut</a>. This forge is just great, and I migrated everything I do to it, but it has a large surface area (and limited manpower) and I tend to easily spot (mostly small) bugs in software I use a lot, so I generate a lot of work for myself :-)</p> The (mostly) Go and Python code base is really solid, and I've been enjoying working on it... for the most part. What I have a "grudge" with is the data access frameworks SourceHut uses, that are a bit too magic to my liking, and that cost me a few hours of unnecessary hitting my head against the wall.</p> I'll illustrate this with two recent occurrences.</p> On uppercase vs. lowercase</a> </h1> The first concerns Go and happened when I worked on this ticket</a>. My patch</a> is very short, but I spent a lot of time debugging it earlier versions, because... drumroll... I was using updateSSHKeyLastUsed *metaSSHKey</code> instead of UpdateSSHKeyLastUsed *metaSSHKey</code>.</p> Here</a> is the broken</strong> code excerpt</p> var result struct { </span> updateSSHKeyLastUsed *metaSSHKey `json:"updateSSHKeyLastUsed"` </span>} </span>if err := client.Do(ctx, "", "meta.sr.ht", query, &result); err != nil { </span> return nil, err </span>} </span> </span>if result.UpdateSSHKeyLastUsed == nil { </span> return nil, fmt.Errorf("no key found for fingerprint %v", fingerprint) </span>} </span></code></pre> Even though logs clearly showed that the returned JSON had the right data, I was hitting error the "no key found for fingerprint" error because updateSSHKeyLastUsed</code> was nil</code>.</p> Once I changed the 'u' to 'U' and things started to magically work, the problem became obvious (you have it, don't you? If not, shoot me an email</a> and I'll let you know ;-))</p> To be clear, this is 100% on me, and things do work as expected. What I'm not happy with is that there was no warning at compile time, no warning log at runtime, no nothing.</p> My take-away here is that if you write a framework, there's as much value, if not more, in letting your users know and helping them when they screw up, than there is in the actual functionality.</p> What do you mean "and"?</a> </h1> The second adventure takes us to the Python and SQLAlchemy</a> worlds. It's been (many) years since I last used an ORM framework</a> and I had forgotten how magic they are.</p> When working on this ticket</a> I had the pleasure of trying to write a "join + where clause" query, that was 100% not working like it should :-)</p> The code I was using is essentially doing this, and was returning way too many rows:</p> shared = select(Event.id, </span> func.min(EventProjectAssociation.project_id).label("p_id"), </span> func.min(EventProjectAssociation.project_resource_id).label("r_id"))\ </span> .where(Event.event_type == EventType.external_event and column == id)\ </span> .join(EventProjectAssociation)\ </span> .group_by(Event.id) </span></code></pre> If you spotted the problem, congrats! If you have not (yet), here's the code that does exactly what I need:</p> shared = select(Event.id, </span> func.min(EventProjectAssociation.project_id).label("p_id"), </span> func.min(EventProjectAssociation.project_resource_id).label("r_id"))\ </span> .where(Event.event_type == EventType.external_event)\ </span> .where(column == id)\ </span> .join(EventProjectAssociation)\ </span> .group_by(Event.id) </span></code></pre> I only found the issue when, desperate, I printed the generated query, noticing to my horror that the initial code was generating a SQL WHERE clause without</strong> any reference of the column == id</code> part. That's why I was getting so many rows!</p> Again, this is sort of on me (I'm pretty sure I'll find somewhere that where</code> should only take simple expressions), but nothing anywhere told me I was being stupid. This is Python, everything's dynamic, so I'm less mad about this case than the previous one, but still.</p> My take-away here is the same as before plus the fact that I should really RTFM when writing Python, because nothing will save me from myself...</p> Status Report - Week 37 2025 Unknown — Mon, 15 Sep 2025 00:00:00 +0000 Real life</a> </h1> It's been a pretty quiet week. I caught a cold and did not feel super good for a couple of days, but I still managed to remove some items from my "obligatory tasks backlog", e.g. worked on my taxes :-)</p> I finished Mikko Hypponen's If it's smart, it's vulnerable</a>, that was very entertaining. I particularly liked the story about how NotPetya</a> impacted Maersk and how their Active Directory could be saved because one node in Nigeria was offline during the incident, and ended up being the one and only backup they had...</p> I barely ran, with only 18.22km this week.</p> Finally, I have not been able to study that much for GeoGuessr</a>, but I definitely played more than usual and easily remained in the Gold II duels division. I feel that I should be able to go to Gold I within a month; fingers crossed! I finished the week with a ELO of 389 and my "Daily challenge streak" is at 23 days.</p> Open source</a> </h1> On the SourceHut</a> front, I started to work on an existing ticket</a> that's been bothering me for a while: "duplicate" mailing list events in my Activity</a> page due to some lists being part of multiple projects. I submitted a "bad" patch</a>, and follow-up with a new one that will be nicer, but is a bit of PITA to write because the data model is not ideal, and we obviously need to cope with existing data. I identified a deletion bug</a> along the way, that I'll handle once I'm done with this one, as well as yet another configuration issue in sr.ht-container-compose</a>, that I fixed here</a>.</p> On the GCC</a> side, I resurrected an old patch</a> and started to work on the review comments, but did not go very far... maybe next week? :-)</p> Status Report - Week 36 2025 Unknown — Sun, 07 Sep 2025 00:00:00 +0000 Real life</a> </h1> On Tuesday, the kids finally started school again, and we're getting back to our normal routine: prepare lunch boxes, help with homework, etc.</p> We had a 24h+ internet outage at home starting Wednesday morning, and Ziggo claimed that everything was fine on their end. The support lady I had on the phone was OK, but could not do anything more than order me a new Internet box since my current one is allegedly outdated. Anyway, everything came back by magic on Thursday late afternoon... I really suspect that something was wrong on Ziggo's side, and that they either did not tell me or did not know about it; I'd be interested in knowing what kind of monitoring and troubleshooting tools they have... Anyway, we have a new box, that I'll probably setup in the coming days, though I'm not hurry...</p> In terms of reading, I have finished Never go back</a>, and make good progress on How propaganda works</a>. The latter is good but not outstanding, and I might pause it a bit again because I came across Miko Hypponen</a> by chance (I guess I can thank YouTube recommendations for once...), and want to read his If it's smart, it's vulnerable</a>.</p> I bit more running this week, with 28.43km. I'm still doing the vast majority of my runs in Z2, but I snuck in a 40 minute intervals session this week. I think the training is going well overall, and that it's useful to be patient and "run slow to run fast". I'll stick to this strategy for the moment.</p> Finally, I'm trying to be a bit more structured and systematic at GeoGuessr</a>, and want to see how far I can get (disclaimer: I'm starting from very low ;-)). I plan to play a lot more</strong> than I did up to now (I've been basically vibe playing my 20 weekly duels games and that's it, hoping for the best), and we'll see where it takes me. I asked my friend Lumo</a> to come up with a training plan</a> with 2 hours of play per day, and will see if it makes sense and how long I stick to it :-)</p> Open source</a> </h1> The kids being back to school, I have a bit more time to code and it showed; let's hope it was not a fluke and I can continue contributing a bunch!</p> Let's start with SourceHut</a>...</p> I followed-up on last week's work to actually maintain the "Last Used" attribute of SSH keys, and submitted two PRs: one</a> in meta.sr.ht and another</a> in sourcehut-ssh</li> I fixed</a> a bug I noticed while investigating threading issues a couple of weeks ago, due to our inconsistent management of angle brackets in mail headers</li> I fixed</a> an issue in the "Patches" view, that led us to not</strong> render links in some messages</li> I looked a bit into the performance of git operations and identified a little pessimization in the commit path, that I sent a RFC patch</a> for.</li> </ul> In the sr.ht-container-compose</a> area, a bunch of my patches were finally merged:</p> Add support</a> for SMTP ingestion to lists.sr.ht</li> Make hut work again</a></li> Make git fully work</a> and upgrade to Alpine 3.22</a></li> </ul> And finally, I feel the itch to resume working on GCC</a>, and fixed the build on MacOS</a>. I hope to write more meaty patches soon.</p> Status Report - Week 35 2025 Unknown — Mon, 01 Sep 2025 00:00:00 +0000 Real life</a> </h1> This was the last week before the kids go back to school, so I tried to make the best of it; we went to the beach, played bowling, watched YouTube videos together, and had a great day at the Veluwe</a> on Sunday.</p> The weekend was very good for the latter, since both the Ultra Trail du Mont-Blanc</a> and the GeoGuessr World Championship Finals</a> were happening, both with very good coverage on YouTube.</p> I did not find the UTMB men race very interesting (especially after my favourite, François d'Haene, withdrew), but the women race was awesome. I was rooting for Camille Bruyas (I really like how she's always smiling / laughing, even when I'm pretty sure things are not fun), and she finished second, so I was very happy.</p> I'll be writing more about GeoGuessr soon, but the short version is that the competition was outstanding, the players really excellent, and my favourite, Debre</a>, finished second to an impressive Radu C</a>. I was a bit sad, but really blown away by the extraordinary skills of all those players.</p> I did not read that much, just making progress in Lee Child's Never go back</a>.</p> It was a short week in terms of running, with only 14.41km, and a pretty good one GeoGuessr wise, placing 11th in the Gold II duels</a> division.</p> Open source</a> </h1> I continued to work on SourceHut</a>, this time in the area of SSH key management:</p> I made it clearer to users what to expect when they remove a key from their account, via this</a>, this</a> and that</a>.</li> I made both pulling and</em> pushing to git repos work with sr.ht-container-compose</a> via this</a>.</li> I submitted a patch to make the "Last Used" information we show for SSH keys actually correct, via this</a>. There's some discussion on it, and I'll follow-up with a v2 the coming week.</li> </ul> In the aerc</a> area,</p> My last pending patch</a>, detecting new emails for GMail in spite of them not support the \Recent</code> IMAP flag, was merged. It's not exactly great, and I'll probably rework it a bit at some point, but it addresses a gap that's been reported quite a few times recently, so it's a net win.</li> aerc 0.21 was released</a> this week. It's a bit vain, but I was happy to see I was the top contributor in this release, with 41 commits.</li> </ul> Status Report - Week 34 2025 Unknown — Mon, 25 Aug 2025 00:00:00 +0000 Real life</a> </h1> We're slowly getting closer to the end of the vacation period, and focus on chilling and charging the batteries for the year ahead. One friend of Lyla's comes home for one week, but it won't be a lot of work for us, because they're old and autonomous enough to know how to occupy and take care of themselves.</p> We finished watching season 2 of Tracker</a>, that obviously ends with a plot twist... I really like this show, and look forward to season 3 (and maybe reading before that the book it's inspired from, The Never Game</a>).</p> Speaking of reading, I started How propaganda works</a>, so far interesting but a bit heavier / philosophy oriented than that author's more recent work. To balance this, I decided to resume the Jack Reacher series in parallel, with Never go back</a>.</p> Finally, I'm back to running and did 22.87km of pure zone 2. Let's see how long I manage to keep ~80% of my training in this zone, and what the effects are.</p> Open source</a> </h1> I only worked on SourceHut</a> this week, in particular on a threading issue</a> in the mailing list software. It was fun in multiple ways:</p> It's a bit of an Heisenbug</a>, and I love</em> detective work :-) I had to go old school and just read the code and think of what could go wrong.</li> I did not find a real fix, so sent an RFC patch</a> that started a good discussion, got me to read and learn about transactions in Postgres and come up with a real fix</a> that I'm actually comfortable with \o/</li> </ul> Status Report - Week 33 2025 Unknown — Tue, 19 Aug 2025 00:00:00 +0000 Real life</a> </h1> This week my brother and his family, as well as my mother, were still visiting, so our focus was to show them around:</p> Gouda</a></li> The NEMO</a> science museum, really interesting both for kids and grown-ups</li> Delft</a></li> </ul> A slow week in terms of reading, with a single book: Baise-ball à la Baule</a>. It'd been years since I had not read a San Antonio "adventure", and I now remember why: the style is very peculiar, and the stories far from outstanding.</p> Open source</a> </h1> No activity this week due to my family visiting.</p> Status Report - Week 32 2025 Unknown — Mon, 18 Aug 2025 00:00:00 +0000 Real life</a> </h1> We had a few days at home to chill before the arrival of my brother, my sister-in-law, my nephew and niece and my mother (pheeew!) for a 10 day visit. And the same way one usually gets sick right at the start of one's vacation, I decided to get pretty sick right at the start of my second (or third?) phase of vacation :-)</p> We visited the Escher museum</a> in Den Haag (the kids loved it) and enjoyed Scheveningen and its waterfront. One good thing with having people visit is that it reminds me how lucky we are to live where we live.</p> I also managed to do a bit of reading:</p> Caleb's guide</a> by Philip Su was very good, and I'll probably tell Lyla to read it</li> A streetcar named Desire</a> was nice, and while it was clear from the beginning that it would not end well, my guess of how it'd turn up was wrong</li> </ul> No running due to my illness.</p> Open source</a> </h1> After a long hiatus, I was happy to be able to send a few patches \o/</p> This aerc patch</a> brings proper label management for GMail and Proton, and closes the last "gap" there was for my email workflows.</p> I decided to tackle a few more papercuts in SourceHut, in particular</p> The automatic superseding of previous patch versions when receiving a new one (patch</a>)</li> A fix in patch author parsing in certain cases (patch</a>)</li> Stop using 32-bit integers to represent database IDs, a little catastrophe waiting to happen (patch</a>)</li> </ul> Status Report - Week 31 2025 Unknown — Tue, 05 Aug 2025 00:00:00 +0000 Real life</a> </h1> For the first time ever, we went to Gran Paradiso park in Italy. In short, it was gorgeous, and really worth the travel. We did two hikes, the first one completely unplanned (because I naively thought there'd be some sort of network where we hiked, and did not prepare anything at the campsite, where we had wifi...) where we saw a "bouquetin" (!), and the second half planned :-)</p> Here are a few pictures:</p> </p> </p> </p> </p> </p> Unfortunately, it was then time to consider going back to the Netherlands. Fortunately we had a stop at Strasbourg to meet some friends, and enjoyed the city in spite of pouring rain :-)</p> Some more pictures:</p> </p> </p> 11.73km of hiking (525m of elevation).</p> Open source</a> </h1> No work at all; I'm getting a bit itchy though, and look forward to getting back to coding a bit next week!</p> Status Report - Week 30 2025 Unknown — Mon, 28 Jul 2025 00:00:00 +0000 Real life</a> </h1> We continue our great vacation, and eventually leave Chamonix to go closer to Matterhorn, my favourite mountain. Since we were in Zermatt a few years back, I decided to get to it from the Italian side and ... did not go the right valley :-)</p> We still did super nice hikes, including our biggest ever in terms of elevation: 1000m+!</p> Here are a few pictures:</p> </p> </p> </p> </p> </p> </p> I read 2 of the 4 books I bought in Chamonix:</p> SOS Himalaya</a>, that I really liked. I like how it illustrates the tension between individualism ("I want to get up there") and support ("If we don't do anything, they will die") among alpinists.</li> </ul> Le crépuscule des jeux</a>, a very interesting "study" about the upcoming 2030 Winter Olympics in France. Long story short: nobody but a few politicians and the olympic committee want it, and it will be a disaster for the environment and France's (already tight budget).</li> </ul> 25.55km of hiking (1851m of elevation).</p> Open source</a> </h1> No work at all; after all, it's vacation time :-)</p> Status Report - Week 29 2025 Unknown — Mon, 21 Jul 2025 00:00:00 +0000 Real life</a> </h1> We finally reunited with Lyla, who's been away in internships then with friends for 4 weeks, and start our "family vacation".</p> We went straight to Chamonix to enjoy hikes and reasonable temperatures, and set up camp at the Camping de la Mer de Glace</a>. This campsite is still excellent, and while our initial plan was to stay for around 5 days, we did 9 :-)</p> Here are a few pictures from our hikes and visits:</p> </p> </p> </p> The good thing with camping is that there's no TV nor PC, and I was able to read a few books:</p> Norwegian Method</a>, that inspired me to try out this method after the vacation, when I finally get back to seriously running.</li> L'ère de la post-vérité</a>, that I really liked but also depressed me a lot...</li> </ul> No running this week, but 28.53km of hiking (and 1189m of elevation).</p> Open source</a> </h1> No work at all; after all, it's vacation time :-)</p> Status Report - Week 28 2025 Unknown — Mon, 14 Jul 2025 00:00:00 +0000 Real life</a> </h1> After attending my sister in law's birthday part in Clermont-Ferrant, we drove back home for a few days of doing nothing before reuniting with Lyla and starting our "family vacation".</p> The only remotely interesting thing I did is play a bit with ChatGPT to create a stupid page that quizzes me about country trivia, so that I improve my GeoGuessr level. The code</a> is not too bad, and does the job; you can play with it here</a>.</p> In terms of reading, I finished the first part of Bill Gates' autobiography: Source Code</a>, and liked it a lot. I like more and more reading about the history of computing, which probably means I'm getting closer to dinosaur status :-)</p> Polo and I did an uneventful 10km run, that showed me that I really need to get back to training...</p> Open source</a> </h1> I decided to tackle one of the remaining gaps in aerc</a> for me to be able to do everything I need in it: proper label management.</p> I'm happy to say that with this patch</a>, aerc supports (un)labelling messages for Proton and GMail \o/ I plan to write a dedicated blog post about my adventures in labels-in-imap-land.</p> There are still a few papercuts that I plan to address after my vacation.</p> Status Report - Weeks 26 & 27 2025 Unknown — Tue, 08 Jul 2025 00:00:00 +0000 This update combines two weeks since I was mostly on vacation.</p> Real life</a> </h1> The first week started with Anton's school being closed for 3 days because of the NATO Summit in The Hague. At the end of the event, the organizers allowed the public to visit the premises (and we did); it was interesting to see the various meeting / conference / 1:1 rooms, all in a temporary building that was literally built ON the boulevard for the occasion.</p> </p> At the end of the week, I traveled to Paris to attend the Top 14 (French top rugby championship) finale between Stade Toulousain ("my" team) and Union Bordeaux Bègles with my kids.</p> Before picking up Lyla at the station, Anton and I visited the Arc de Triomphe (first time for both of us) and enjoyed the great 360 degree view on the city</p> </p> The rugby game was outstanding: both teams played really well, and the regular 80 minutes were not enough to decide on a winner: for the first time in years, a 20 minute overtime was required! And since Toulouse eventually won, we were super happy \o/</p> </p> In the second week, Lyla went back to her friend's and Anton, Nathalie and I went on a camping trip in Normandy, with a destination in Clermont-Ferrand for my sister-in-law's birthday party.</p> Staying relatively up North allowed us to escape the big heatwave over France, and we visited the American Military Cemetary - a great way to remember the sacrifices of the young soldiers who came to defeat the Nazis.</p> </p> I barely read anything, and did ZERO running :-/</p> Open source</a> </h1> I almost did nothing "work wise", except for a small simplification</a> in aerc.</p> On the SourceHut front, my patch</a> allowing patch submitter to withdraw or update their own patches was finally deployed \o/</p> Status Report - Week 25 2025 Unknown — Mon, 23 Jun 2025 00:00:00 +0000 Real life</a> </h1> This week involved some travel again, this time greener: Clermont-Ferrand => Paris => Lille by train and Lille => Den Haag by bus. It all went very well, with barely any delay, and good wifi.</p> This helped me increase my books_read_count</code> by 3: Je m'en vais</a> (a re-read from 2008 (!), still very good), Les morts ont la parole</a> and Big Bang et au-delà</a>.</p> And I achieved one of my worst weeks in terms of running volume, with a bare 4,24 km :-) Let's blame the fact that I've been a bit sick.</p> Things were a bit better in terms of (e-)sports, with my qualification to the Gold I</a> division at GeoGuessr</a> \o/</p> Open source</a> </h1> A quiet week here as well... I made two commits to aerc: a workaround</a> for Zoho.com, and a crash fix</a>.</p> As a follow-up to my recent breakage of mail archiving in lists.sr.ht, I added support for end-to-end delivery</a> to mailing lists to sr.ht-container-compose</a>.</p> I leveraged this to fully test</a> and submit the (hopefully) last revision of the patch</a> to allow patch submitters to withdraw their own patches.</p> Goodbye WorkMail, hello Proton Mail Unknown — Sun, 22 Jun 2025 00:00:00 +0000 When we moved to the Netherlands and I started working for AWS WorkMail</a> in May 2016, I migrated my personal email to this service (I'm a big fan of eating my own dogfood!), and I have been happy ever since: it is relatively cheap, definitely does the job for me [1], and knowing how it works inside always felt nice.</p> However, I recently looked at my recurring "IT costs" and was reminded that small amounts can add up: between WorkMail, LastPass and ExpressVPN, I'm paying 18.6 euros every month. This is not the end of the world [2] but not cheap either, so it was worth looking at what's out there.</p> Proton</a> quickly became a top contender, partly because of its characteristics (who wouldn't want privacy?!) but more importantly because its Proton Unlimited</a> bundle provides all I need at a very competitive price [3]: email, password manager, VPN and Drive (and more stuff I'll probably never use).</p> Migrating data</a> </h1> I will not talk about migrating LastPass</a> data to ProtonPass</a> given how trivial it was</a>, and focus on my mail data.</p> I have around 2GB worth of emails, and migrating them to Proton requires</p> Changing DNS records to point my domain to Proton</li> Copying all existing emails to Proton</li> </ol> The first point is very simple thanks to a clear and intuitive UI</a>; I felt surprisingly safe clicking around in Route53</a>, and particularly appreciated the almost instant feedback about my actions.</p> And copying data over turned out to be super easy as well: Proton's Easy Switch</a> tool took its time (around 12 hours) but imported everything with no issue. I really liked that the import is idempotent [4], which gives peace of mind.</p> Using the various products</a> </h1> I migrated everything one month ago, and I'm super happy with my decision.</p> Proton Mail</a> </h2> IMAP access</a> </h3> To maintain its security posture, Proton does not provide native IMAP and SMTP access, and instead requires running a bridge</a> on your computer. It interacts with Proton Mail servers, taking care of all encryption aspects, and exposes IMAP and SMTP end-points on localhost [5]. These are the ones your point your favourite mail client to.</p> Not only does it work really well, but it's also a great quality-of-life improvement when using clients such as aerc</a> that only have a minimal local cache, and interact a lot</strong> with the IMAP server. The bridge completely hides the latency (the server is literally your computer), and ensures a good user experience even when offline or on a flaky network.</p> There are however tiny drawbacks:</p> If you access your mailbox using IMAP from 5 different computers like I do, each will have a full local copy of it (this is good for data resiliency though ;-))</p> </li> The bridge and the import tool work such that emails are sorted by "arrival in the mailbox", not actual receive date. This distinction would not matter if the import did not start with the most recently received emails and moved towards the past :-)</p> </li> </ul> WebApp</a> </h3> I barely use it, but the Proton WebApp is nice: I don't miss any capability compared to GMail, and it's roughly 20 years ahead of the WorkMail WebApp.</p> iPhone App</a> </h3> Having to use a dedicated app instead of the stock Mail iPhone app did not make me happy at first, but I quickly got used to it and must admit that the Proton Mail app is much better (especially in terms of mail tagging/labelling, which I do a lot), and I would not go back even if I could.</p> Proton Pass</a> </h2> The move from LastPass is a clear improvement: all the basic functionalities are there, and I really love the aliases functionality</a>: with the Unlimited plan, you can generate as many aliases as you want, and you don't need to communicate your actual email address to anybody. Proton automagically forwards any email sent to an alias to your mailbox, and you can protect your email address with NO usability friction.</p> Proton VPN</a> </h2> While I was barely using ExpressVPN, I'm using Proton VPN on all my devices. Performance is really good, and apart from one or two apps who refuse to work when it's on on my phone (I'm looking at you LCL</a>), it works extremely well. And I love looking at the number of ads and trackers that have been blocked :-)</p> [1] I'm far from a power user: as long as I can access my mailbox with IMAP and on my phone, I'm happy.</p> [2] I wish I was only spending this for entertainment platforms; I should look at this next, but my kids might not agree :-)</p> [3] As I write this post, 9.99 euros per month when paid annually.</p> [4] Strictly speaking, it's not since each imported message gets labelled with a label identifying the import, but it's a detail.</p> [5] The bridge's security model</a> is a good read.</p> Status Report - Week 24 2025 Unknown — Mon, 16 Jun 2025 00:00:00 +0000 Real life</a> </h1> This week is the first in a series of four "logistically challenging" ones. Lyla finished her school year, but she's also starting her first one week internship at my brother's in Clermont-Ferrand. We travelled for most of the day Saturday; it was a bit long but uneventful. I'm writing this in the train back to the North, with a stop in Lille on Tuesday.</p> One advantage of (long) travels is that they're an opportunity to make progress on my (very long) reading list. This week I crossed out Snap</a>, a typical airport novel you enjoy but forget after 5 minutes, and Working in public</a>, particularly interesting as I'm asking myself many questions about what I want to do professionally in the future.</p> In Clermont-Ferrand, we went to Les Nuées Ardentes</a>, an event that the university organizes every year where various departments make nice demos about their field. We solved a cool "cryptography escape game", measured the propagation speed of waves, or saw how carbon dioxyde makes water more acidic. It's really nice to see passionate people take (personal) time to freely share their passion with others.</p> Open source</a> </h1> The main event this week is that one</a> of my patches to lists.sr.ht broke delivery to mailing lists :-( Fortunately I happened to see that the patch had been integrated and soon after tried to submit another patch that did not seem to be received... so I let folks know on IRC that things did not seem to work, and Drew DeVault quickly deployed a fix</a>.</p> I'm still pondering whether writing a COE</a> for this would be of any value, but I've already taken the action to add support for full inbound email processing in sr.ht-container-compose (patch</a>). This will allow to properly end-to-end-test changes to lists.sr.ht, instead of using a partial workaround based on imports (yes, I would have spotted the issue with my patch with this...).</p> Robin Jarry also made the inbound delivery flow more robust to archival errors</a> so all in all my mistake triggered good improvements to the service.</p> And finally on the aerc front, following a report</a> that marking messages read felt slower with master than 0.20, I wrote down a set of nominal scenarios</a> involving read/unread status. Thanks to this (super boring) exercise, I identified three issues with my recent commits ( 1</a>, 2</a> and 3</a>), and the root cause for a long-standing issue with Office365 (fix</a>).</p> I also root-caused the reported slowness to a Zoho specific issue, that I'll try to workaround this week.</p> Status Report - Week 23 2025 Unknown — Tue, 10 Jun 2025 00:00:00 +0000 So it looks like the idea of writing a weekly update will be more than a one-off \o/</p> Real life</a> </h1> There's a lot to love about the month of May: we're just out of a 4 day weekend, and have a 3 day one coming :-)</p> On the flip side, Nathalie's students start freaking out about their coming exams and ask for more sessions [1]. It's good for her business, but logistically challenging. Fortunately it'll be over in a few days.</p> We watched I Daniel Blake</a>, a very good-but-also-super-sad movie. I like to think that my kids are aware of how lucky we are, still it's worth having refreshers from time to time.</p> At age 43, I finally read my first book by Gabriel García Márquez, 100 years of solitude</a> and it was awesome. Now the problem is that my "I need to read those" list, already very long, just got longer :-)</p> And finally, a bit more running this week: 26.5km. Nowhere near where I should be given that I'm, well, not really working, but it's better than last week, so progress.</p> Open source</a> </h1> The "pipe situation" that I talked about here</a> continued with a twist: ~rjarry</a> found a much better fix</a> to the initial problem. The time I spent investigating this was not lost however - I had fun and learnt - and that's one nasty bug less for aerc users \o/</p> I also submitted a patch</a> to configure a duration after which a message being previewed in the message list split</a> is considered as read. This addresses the friction of still having to explicitly open or mark read a message even if you've fully read it in the split view.</p> And I dug myself into a small rabbit hole on the SourceHut code front when I wanted to highlight a few lines on git.sr.ht</a>'s file view using something like aerc.1.scd#L725-734</code>, and noticed that it did not work. To be able to investigate the issue, I had to submit a few patches to sr.ht-container-compose</a> to make the git server work ( this</a>, this</a>, this</a> and that</a>), and was finally able to submit a fix</a> :-)</p> [1] One would hope it's "I know everything, but would like to refine a bit X" instead of "Oh sh... I haven't done anything in the last 5 years, please save me" sessions, but it's not :'(</p> Status Report - Week 22 2025 Unknown — Mon, 02 Jun 2025 00:00:00 +0000 Moving forward, I'll (try to) write a weekly status report to document what I did the previous week. This is NOT for Elon [1], but primarily for me and to mitigate my super power of quickly forgetting things.</p> Real life</a> </h1> We had a long weekend here in the Netherlands - Thursday to Sunday - but the weather was not great so we did not do much. The kids start to be tired by the long school year, and being able to rest doing nothing was much appreciated.</p> We continue watching the Tracker</a> TV series (I like it a lot!), and watched a movie about Tonya Harding, I Tonya</a>, that I found very depressing.</p> On the reading front, I finished Ha-Joon Chang´s very good Economics: the user's guide</a>, read Daniel Stenberg's OK uncurled</a> and Laure Noualhat's frustrating Le nucléaire va ruiner la France</a>.</p> And finally on the running front, there was really not much running - 12.74 km :-/ It will be better next week.</p> Open source [2]</a> </h1> The most noticeable event is the unexpected and very nice shout-out that Drew DeVault made in his What's cooking on SourceHut? Q2 2025</a> blog post. At least for a few days, simartin.dev</a> was accessed by some people other than myself or my family :-D</p> I spend quite some time on an aerc investigation, that got me to learn more about Unix pipes, (finally) how to debug Go programs with an actual debugger and not print statements, and to dig into some of aerc's dependencies (that I had to patch</a>). More can be found in the blog post</a> I wrote about the adventure.</p> I also submitted a patch</a> to allow aerc users to prevent mail deletions from a folder that's not the Junk nor Trash folder. That's because unlike clients such as GMail or MailMate, "Delete" in aerc means delete, not "Move to Trash", and I would not be surprised if many users have found out the distinction the hard way.</p> And since Drew mentioned it as done in the aforementioned blog post, I finalized the patch</a> to allow patch submitters to withdraw their own patches on SourceHut. I also iterated on the one</a> to show a meaningful activity date for SourceHut mailing lists.</p> Next week, I hope to find some time to revive some GCC patches ( this</a> and that</a>) for which I got feedback that I need to act on.</p> [1] Who seems to have made the world aware of the common practice in the professional world of asking people to regularly give a status of their on-going work - something that I did for years at Amadeus</a>.</p> [2] I was thinking about calling this section "Work", but since I can decide to not do anything for weeks with no consequences, and I'm not being paid for any of this, it's not really work :-)</p> Digging myself into (and out of) a rabbit hole^W pipe Unknown — Thu, 29 May 2025 00:00:00 +0000 I've been hacking on aerc</a> lately, and I'd like to talk about a very recent issue I've been fighting with. It's been super frustrating, but I also learnt a lot, and eventually came out "victorious"... so in the end, it was fun.</p> If you talk too much, I'll shut you off!</a> </h1> You can almost do anything with aerc: it provides generic mechanisms and is highly configurable, so the sky is the limit. If you want to see an email's raw data, you can just pipe its content into less</code> using :pipe -s less</code> and inspect the data in the tab that this opens. And if you have selected several emails, it will also work - the messages will be combined into a Mbox</a>.</p> Excited about this, and because it seems I like to f... around with software, I selected a bunch of messages, pipe -s less</code>-d them and... nothing. Not nothing like "it does not do anything", but rather like "I am frozen, and your only option is to kill -9</code> me".</p> A big believer that "a problem reproduced is half solved", I started to read the code</a> in charge of the pipe command, thinking I'd quickly get to the bottom of the problem. How naive! :-)</p> pipes for dummies</a> </h1> Needless to say, everything looked fine in the code, so I started debugging (aerc is written in Go</a>, and I'm still early in my journey with this language, so "debugging" means "printf debugging" here), and was going nowhere... so I decided to invest a bit of time and teach myself how to debug Go programs [1] using delve</a>.</p> It turns out that aerc was blocking on a call to io.Copy</code></a>, and a quick search taught me that (Unix) pipes don't have an infinite buffer (duh!), and that if you try to write into a pipe with a full buffer, you'll block until something reads data from the other end.</p> Since aerc's:pipe</code> is structured such that a Go routine writes data into the pipe and then signals another Go routine to start reading, the idea of the fix is trivial: (1) figure out what the pipe limit buffer is, and (2) make sure that you don't write more than that.</p> How much it too much?</a> </h2> The first point already is not that straight-forward: like most low-level APIs, the answer is that it depends :-) Fortunately, I found a great answer</a> on Stack Exchange.</p> I naively hoped that Go would expose an API to find out the exact limit, so that we can KNOW and not guess, but it's unfortunately not the case... So I convinced myself that I would not be too wrong in working with a limit of 65535 bytes [2], wrote some code, and could quickly start testing.</p> All was good when piping a single message, but when piping several, I noticed that the warning I added to notify the user that data was truncated due to pipe limitations did not always show.</p> I tried a few changes, but I was going nowhere, and got frustrated by the fact that to test every change, I had to compile, open aerc, load my mailbox, select N messages and pipe -s less</code> them... and see that the problem was still there.</p> Interlude on testing</a> </h1> aerc has a lot of unit tests, but only a few for UI aspects. I decided to bite the bullet and write some for the :pipe</code> command. As expected, it was not fun, but it proved crucial for the rest of this adventure and my sanity.</p> So yes, some tests are a pain to write, but don't chicken out, and write them. It will pay (lots of) dividends down the line.</p> Counting (bytes) when someone is lying</a> </h1> So now I had four things: a test demonstrating that indeed, there was a bug with "Mbox mode", my two eyes (to cry :-)), and my mouth (to curse).</p> After (too many) trials and errors, a few runs to think about something else, and lots of looking at the code trying to understand why it'd spit things like invalid write result</code> or short write</code>, I asked myself the question that I should have asked from the start [3]: "what if something is lying?".</p> From there things got easy: I did not look only at aerc code, but also at that of its dependencies, in particular go-mbox</a>. I quickly spotted suspicious code, and an old ticket</a> reporting a pretty similar issue.</p> After submitting a fix</a> for that issue, verifying that everything [4] is fine with that fix, and adapting the aerc code and tests to live until a new version of go-mbox is released and integrated, I was finally</strong> able to submit a patch</a> fixing the freeze \o/</p> Concluding remarks</a> </h1> This investigation took place over 3 days, and while it was definitely not my longest [5], I'll remember it for a few reasons:</p> It got me to dive into lowish level stuff, something I love but had not done in years,</li> It forced me to look into the debugger landscape for Go,</li> In the end, I solved the issue, improving both aerc and go-mbox</li> </ul> June 3rd, 2025 update</a> </h1> The patch I proposed was eventually discarded, since ~rjarry</a> found a better one</a> that does not involve any truncation. The majority of this post remains valid.</p> [1] It was literally an investment of a few minutes; the go ecosystem is rich and well documented.</p> [2] Spoiler alert: I was naive, again :-)</p> [3] It's always like that, isn't it?</p> [4] go-mbox's Writer still lies to its users if/when it replaces "\r\n" by "\n" and it's not as easy to fix; I might or might not try to fix it in the future.</p> [5] The longest was clearly at Amadeus</a>, when a new version of our home-made service bus would crash within 2 minutes to 1 hour, in Production only. It took more than two months (not full time, thankfully) to find an (old) race condition due to a mishandled double (or was it triple? ;-)) negation.</p> SourceHut, or how I learned to love^W tolerate web development Unknown — Tue, 15 Apr 2025 00:00:00 +0000 Since I decided to start using aerc</a> and it's hosted on SourceHut</a>, I discovered this platform in early March, and quickly became a big fan.</p> I'm happily paying my (optional</a>) 5 dollars a month, and migrated a lot of my day-to-day stuff (this website, my personal todo management system and various code repositories) to the platform.</p> In the last 25 years, I had only used 2 other "forges" - SourceForge</a> and GitHub</a> - and the huge</strong> advantage of SourceHut over those is that it's fully open source. So if you're annoyed by a bug or have an improvement idea, you can just git clone</code>, fire up Sublime</a> and off you go :-)</p> In this post, I'll go through a few things I contributed (not all are accepted yet).</p> Testing infrastructure contributions</a> </h2> The SourceHut software can be deployed</a> by anybody on their own hardware, and since I wanted to make code contributions and test them before submission, I slowly started to follow the deployment and hacking guides.</p> It's not rocket science, but not super fun either, so I did not make much progress, until I came across this repository</a>, that deploys an instance with most services using Docker Compose, and gave it a try.</p> The bad news is that it did not really work (at least for me), but the good one is that, again, it's open source, so you know what happened next (git</code>, Sublime and all :-))</p> Properly setup tokens => patch</a></li> Make services actually talk to one another => patch</a></li> Make sshd happy => patch</a></li> </ul> Finally able to deploy and play with the code, I moved to more visible stuff, leaving my familiar backend and infrastructure worlds towards... web development!</p> Patch management improvements</a> </h2> SourceHut's patch management philosophy is similar to that of GCC or the Linux Kernel: you send patches by email, review patches by email, and (probably, I have not done it yet) merge patches by email.</p> All this communication is done through a mailing list (e.g. this</a> for aerc), and it it's hosted on SourceHut, the UI provides a nice "patch oriented" view (this</a> for aerc).</p> Each patch has a status, that can only be changed by list moderators, which is great for control, but sucks for folks like me who have borderline OCD and cringe when they send multiple revisions of a patch, see that they're all "PROPOSED" and cannot</strong> mark previous obsolete revisions as "SUPERSEDED" on their own.</p> </p> That said, it's open source, so you know what happened next (git</code>, Sublime and all :-))</p> Allow patch submitters to update their own patches => patch</a></li> </ul> Search improvement</a> </h2> SourceHut does not have some of the bells and whistles that GitHub has, which is a good thing, but I sort of miss GitHub's contribution graph</a>.</p> At some point, I want to be able to show a similar graph on this website for all</strong> my open source contributions, not only those on GitHub, and I am therefore looking for reliable ways to find contributions on the various platforms.</p> For SourceHut, mailing lists seem to be the right way, especially since they support the "from:" search syntax, so all is good... except that not :-)</p> That said, it's open source, so you know what happened next (git</code>, Sublime and all :-))</p> "from:me" does not find messages generated by the patch UI => i_still_need_to_fix_this</em></li> Don't let the existence of random users poison from: searches => patch</a></li> </ul> Having fun with aerc Unknown — Sun, 06 Apr 2025 00:00:00 +0000 I recently discovered the aerc</a> mail client and, appealed by its "First class support for working with git and email" [1], decided to try it out.</p> The quick summary of my experience is that (1) this client is great, but (2) there are a few papercuts that will make folks with slight OCD (e.g. yours truly) cringe.</p> The great thing with free software is that you can go fix the bugs that annoy you, and I embarked in a journey to get make my experience with aerc even better.</p> First experience with Go</a> </h2> Call me a dinosaur / snob / boring, but I have happily ignored the "Go</a> is great, you should write in Go" for the last 18 years [2], and was not missing out.</p> However, aerc</a> being written in Go, I had to take the plunge two weeks ago, go a bit deeper than "Yeah I know Go, it's the thing from Google" and actually read and write some.</p> The jury is still out, but my early thoughts are that the language is pretty good, easy to understand and write (as advertised), and has a great compiler (I need to confirm but it looks like it operates like GCC does with -Werror</code>, and I'm a big fan) and a solid ecosystem.</p> I had a WTF?! moment when I understood that they use upper vs. lower-case to encode symbol visibility (seriously, who does that? why not use whitespaces to delimit blocks while we're at it? ;-p)), but I'm past this now and I am fine with writing Go.</p> Who cares about unread counters?</a> </h2> So why did I (temporarily) leave my beloved C++ to do Go? Primarily because I quickly noticed a few things when using IMAP to access my (AWS WorkMail) mailbox, especially around unread counters.</p> I like to have no unread message, so it's important that unread counters are correct; what I particularly don't want is to be told a folder contains some unread message that I cannot find... because there's none! This happened quite regularly.</p> The IMAP code in aerc is pretty good, and relatively easy to follow, so I took to fixing those annoyances myself. I still have one PR that needs merging (commits</a>), but I'm now at a place where I cannot "trivially" get those counters out-of-sync anymore, a big win. I'll admit that I have broken a few things along the way, but the aerc developer community has been super nice and understanding... and I like to think that I have also been very reactive.</p> I'll very probably write a bit more about aerc and IMAP, in particular the aspects of the protocol that I discovered and triggered a few WTF!? on my part.</p> Until then, the TL;DR is that (1) aerc is great and really worth looking at, especially if you're a "terminal person", and (2) free software is really nice, because you can get to fix what you don't like and the community is very welcoming.</p> --</p> [1] I spend quite some time using git-send-email for my contributions to GCC</a>.</p> [2] While writing this, I still thought Go was "new". I guess time flies...</p> Et si on se mettait à bloguer 25 après tout le monde ? Unknown — Wed, 02 Apr 2025 00:00:00 +0000 Ayant récemment découvert SourceHut</a> (plus de détails dans un billet futur), son hosting de pages web simples, et des logiciels de génération de sites statiques tels que Zola</a>, me voilà qui démarre un blog, 25 ans après tout le monde.</p> Je pense y parler d'informatique, d'éducation, de géographie et de tout ce qui va m'intéresser au fil du temps. Autant dire que ça va être éclectique et d'une valeur limitée [1] mais ça me fait envie donc... c'est parti !</p> [1] Au-delà de me faire rigoler (et honte) dans 20 ans, comme le fait le site que j'avais quand j'étais en école d'ingénieur (merci archive.org</a> [2]).</p> [2] En parlant d'archive.org</a>, donnez-leur quelques sous si vous pouvez vous le permettre - leur service est extrêmement important pour le monde, et beaucoup essaient de les descendre...</p>