David Marcelis

@gitbutler I'd love that for private repositories! I still want to re organize commits without PRs. Is this available in the GUI as well?

we are testing a "skip the PR" (a.k.a. merge to main) workflow support. if you wanna try it early, check out the latest nightly and try the `but land <branch>` command. and do share any feedback you have! gitbutler.com/nightly

@rovarma Would be very interested!

Related: I die a little bit inside whenever I see new devtools being launched with perpetual pricing. Been thinking about writing a blog post about the commercial realities of devtools, but I don’t know if people are ready to hear the truth :-)

@vkrajacic Cool! That makes sense, as a File Explorer replacement you'd need to support all file systems. And I'm guessing ReadDirectoryChanges for updates after indexing?

@stardidi_tk It doesn't yet. It uses FindFirst/Next and NtQueryDirectoryFile (where it can) for now. I do plan to add direct MFT loading at some point, but only as an optimization path, since that works on NTFS only.

@vkrajacic Does File Pilot rely on anything NTFS specific (MFT/GPT, USN, OpenFileByID)? Curious if it would work on top of a VFS

Errr, GPT obviously shouldn't be there 😆

@chacon Thanks! Managed to catch the livestream on YT

@stardidi_tk There is a google meet, so probably: github.com/jj-vcs/jj/wiki…

I'm weirdly excited about attending JJCon tomorrow. It has vibes of the GitTogethers at Google 15 years ago - just a bunch of nerds nerding about version control.

@ForrestTheWoods @rovarma That would an amazing next project for Superluminal ;)

Is there really not a good way to profile mixed Python / C++ code? Or profile Python memory allocation? Everything I’ve found seems really really bad.

@chacon @gitbutler Had no idea this was happening in the city I work in! Signed up, would love to have a chat about pushing the limits of source control!

A reminder that @gitbutler will be at the Open Source Summit EU next week to participate in the Git Mini Summit. We're buying beers for everyone after, so if you're in Amsterdam, come and join us! #git-mini-summit-2025" target="_blank" rel="nofollow noopener">events.linuxfoundation.org/open-source-su…

@ForrestTheWoods 100% agreed. Though I think we can do vastly better still. Do you use Araxis' option to align all unchanged texted by inserting extra empty space? It sounds nice, but I find I just need the full untouched documents left and right to properly understand the diff

Maybe(?) unpopular opinion: I utterly despise single column diff views. “git diff” is poopy poop. Two column diff view is *radically* superior. Infinitely easier to read and grok. Been using Araxis Merge for almost 20 years. Single column diff is trash. Am I wrong?

@mold_time Cool! It spawned from a comment my mother made that it's really hard to loose weight if you're obese, which made me think that looking at data from children might show trends with lower latency than adults. (Although there are probably a million reasons why that's inaccurate)

@stardidi_tk nice find! we will take a look :)

@mold_time Possibly of interest to you: Germany reduced child obesity by 0.75x in the last 20 years, while the Netherlands' increased by 1.6x. In the general population their trend has been parallel for decades and they are fairly similar countries. It intrigued me :)

@__silent_ I am 98% sure we just hit the exact same bug. Some critical code that had been fine for 6+ years suddenly broke with seemingly corrupted memory. Thank you for writing this! We'd started considering something with W11, but this probably shortcut days of investigation

New blog post! A bug in GTA San Andreas lay dormant for over 20 years, until unrelated changes in Windows 11 24H2 triggered it. This is a deep dive into how a simple coding mistake erased all seaplanes from the game and made them completely unusable. cookieplmonster.github.io/2025/04/23/gta…

@TheGingerBill @rfleury Could you explain that? Why would anyone not want to be able to debug crashes in their shipped executable? You don't want ship the debug info, but that's what separate files / stripping is for.

@rfleury So the problem I am thinking about is what people expect. When building a "release" build, people don't want debug info by default.

Debug info generation should be on in compilers by default.

@DefenceForceOrg @ForrestTheWoods @molecularmusing @atoi6664 @xdfdev Oh man... I am so shielded from the regular world. We don't use msbuild, and we have 0 tolerance for bugs like that. Apart from occasional new issues, rebuilds and cleans are basically nonexistent

@stardidi_tk @ForrestTheWoods @molecularmusing @atoi6664 @xdfdev The thing that would scare me with that is that typically Visual Studio is not very good at staying clean, like sometimes pressing "stop build" would end up with a bunch of corrupted pch, pdbs, obj or lib files, then it would fail on the next build requiring manual deletion

Want to see some cursed C++ code that makes the MSVC toolchain generate a mangled name which cannot be demangled by any tool? Anonymous namespaces, another cursed feature in the context of Live++.

@DefenceForceOrg @ForrestTheWoods @molecularmusing @atoi6664 @xdfdev Oh for sure! I think we even ran into issues around 1GB where sometimes linking would suddenly take 20 MINUTES with the PDB server. Never again 🤣

@stardidi_tk @ForrestTheWoods @molecularmusing @atoi6664 @xdfdev Forgot about that one: PDB size, when the symbol server was failing because a PDB was larger than 2 gigabyte...

@ForrestTheWoods @molecularmusing @atoi6664 @DefenceForceOrg @xdfdev But I think the only way it's gonna be fast at that scale is by having a linker that just stays in memory and can update what has changed instead of having to re-parse 10GB of linker input

@ForrestTheWoods @molecularmusing @atoi6664 @DefenceForceOrg @xdfdev I think we now use lld combined with post-processing of the Z7 output to insert hashes so the linker can more easily de-duplicate debug symbols. It's decently fast (at roughly Unreal Engine scale). From memory around 20s for a 200MB exe and 1GB pdb

@molecularmusing @ForrestTheWoods @atoi6664 @DefenceForceOrg @xdfdev But as linkers have gotten faster, we should re-evaluate that trade-off. The sweet spot for unity size has probably moved. And as others mentioned, the biggest gain is probably to clean up include hierarchy. A rough estimate showed that could save 90% of frontend time...

@molecularmusing @ForrestTheWoods @atoi6664 @DefenceForceOrg @xdfdev Right! In a distant past we used to have link times of around 2 minutes. Unity builds brought that down to 1 minute by reducing duplicate obj data. Having a single compilation unit go from 1s to 10s is still a net gain compared to the 60s of saved linking. (Numbers approximate)