Markus Stange

At Mozilla, working on Firefox for macOS, and on the Firefox profiler. he/him
@[email protected]

[Markus Stange @markusstange]

@jasonlaster11 Nice! Are you planning to make it asynchronous / chunked, too? I’m interested in something similar for the profiler, which currently uses a full EditorView. See https://t.co/6buioFN7Eb

[Markus Stange @markusstange]

@timmisiak @wesleywiser Wha…?!
I thought a Bytes was just a few usize values, and the actual contents were stored out of line. Very confused.

[Markus Stange @markusstange]

@stevemk14ebr @timmisiak Oh cool, you're using https://t.co/k258wKq0NR ! Very nice.

[Markus Stange @markusstange]

@timmisiak I'm definitely interested in contributions! I'd even be happy to pass off ownership completely, if that's something you're interested in.

[Markus Stange @markusstange]

@timmisiak @Gankra_ I’m making a profiler and I would like it to work out of the box for people who already have _NT_SYMBOL_PATH set just the way they need it.

[Markus Stange @markusstange]

@jnadeau As someone who always hears the word “Preferences” in Duke Nukem’s voice in my head, it made me very happy to see you use TextEdit as the example app for the rename.

https://t.co/zr9A5rdTtk

[Markus Stange @markusstange]

@Gankra_ Also, they came at exactly the right time and allowed me to solve a real problem: My frame pointer stack walker was producing bad stacks in arm64 leaf functions. Now my stacks are pristine.

https://t.co/Yq3lTjyhiH

[Markus Stange @markusstange]

@Gankra_ The docs were so compelling that they made me want to write my own implementation for fun, regardless of how good the original implementation was. I didn’t really look at your code, I just saw that it wasn’t standalone, and that gave me enough of an excuse to make a new one.

[Markus Stange @markusstange]

@NewAgeDerpDerp @gewt It is, the cursor stuff was fixed both in the browser and in the OS. This must be something else.

[Markus Stange @markusstange]

@iximeow I have a wip! Just started a week ago. https://t.co/wII91EcC24

[Markus Stange @markusstange]

@kubkon @ziglang I'd be very curious to read about the entitlement solution, once you figure it out.
I'm currently using a DYLD_INSERT_LIBRARIES hack to smuggle out the mach port from the child process, but it only works in limited cases. https://t.co/ABmCa78u1y

[Markus Stange @markusstange]

@simevidas Ah, yeah, the overscroll background. But also the "compact tab layout" toolbar coloring.

[Markus Stange @markusstange]

@simevidas Heh, that is indeed rather faint.

I was wondering why they didn't set a bright root background for the blog, so that Safari's scrollbar heuristic would work properly, but I suspect that would break the Safari toolbar coloring.

[Markus Stange @markusstange]

@badlogicgames @mitsuhiko @molecularmusing Can you try out https://t.co/BLjpdKDyPc ? I’ve given jrmuizel’s proof of concept a proper API, improved performance and added lots of caching. It should be super fast.

[Markus Stange @markusstange]

@mitsuhiko Thanks for the kind words!
I'm currently working towards a proper release with a bit more polish, better docs, and a new name.

[Markus Stange @markusstange]

@Gankra_ @jckarter @pcwalton @mitsuhiko Agreed. The frameless opcode is legitimately used for leaf functions.

[Markus Stange @markusstange]

@Gankra_ @jckarter @pcwalton @mitsuhiko Which is funny to me, because there are plenty of free bits in the arm64 "frameless" __unwind_info opcode, and those bits could have been used for information about where the registers are on the stack. This was clearly deliberate.

[Markus Stange @markusstange]

@Gankra_ @jckarter @pcwalton @mitsuhiko With arm64, they've gone all-in on frame pointers even in the compact CFI. All non-leaf functions without frame pointers get hit with a "use dwarf unwinding" opcode + __eh_frame CFI.
Enabling frame pointers in arm64 Firefox saved us 8MB of unwind data. https://t.co/F5LVXiOjX3