Exactly. There's a reason why 'no commercial use' licenses are generally best avoided: defining commercial use is hard. The problem for the FAA is that, traditionally, it's pretty easy with an aeroplane: if someone is paying you to fly the plane, it's commercial, otherwise it isn't. The distinction makes sense because you want tighter regulation on pilots who are going to fly with passengers, or for those passengers to definitely know that they're flying with a hobbyist at their own risk. For drones, it makes a lot less sense.

Well, fuck you Slashcode! Apparently I was using too many junk characters by having the temerity to post code snippets. Posting lots of mathematics also triggers it. Remember when this place used to be for nerds? Rather than try to work around the filter, I have placed the contents of my post here: http://pastebin.com/HtQXTnX0

I don't think I understand what you think you're trying to do. You can't make a cache flush a line that you're modifying with an atomic operation to RAM, because atomic ops require the value to be in cache. Given an n-way set associative cache, however, you can typically force cache flushes (without requiring special cache flush instructions) by writing N+1 values at cache-line offsets (e.g. at address X, X+64, X+128,...) repeatedly. This probably wouldn't trigger the rowhammer issues though, because it's up to the CPU which row it evicts each time and you'd end up repeatedly stalling on loads without bashing a single DRAM line. You might be able to do something similar with the nontemporal store instructions that Intel added in recent generations of processor...

It's not just about freedom, it's also about economics. Copying software (and music) is trivial. Writing software is hard (well, writing good software is, at least). With free software, you don't charge for copying, but you often do charge for writing the software in the first place. And, because of the relevant licenses, there's a large body of code that you can charge for fixing / extending / customising.

Huh? You want on-stack arrays, std::array gives you an on-stack array. You want to pass them to a function, a template function that takes std::array references would work fine.

The most ridiculous part of the summary is the part where she thinks that an Internet-connected system is secure if no one has physical access to it...

I think that the single USB-C port is going to be pretty annoying for the people who buy one this year, but it is likely to have the same effect as making the iMac USB only: provide a sufficiently large demand for USB-C devices that it makes sense for peripheral manufacturers to produce docking stations, displays that can provide power over USB-C, and so on. In a couple of years, I expect that there will be enough devices on the market handle breakout from a single USB-C connector that people buying laptops won't have a problem with it.

However, just like the original iMac, there's going to be a lead time where the only peripherals that you can connect reliably are (expensive) Apple-branded ones. If you remember the iMac launch, you'll recall that about a year later computer stores were full of USB stuff all in the same sort of translucent coloured plastic as the iMac to encourage iMac users to buy them, but a year after that the vast majority were bought to plug into non-Apple machines.

Not with the latest version, Ethernet was gone several revisions ago. Most people really don't need it. I generally only use it at work, and leave the dongle connected to the Ethernet cable, so it's no harder to plug in than connecting the cable itself.

The idea is that you'll just have one cable coming from your monitor, providing power and a USB hub. Currently you can only get that with Thunderbolt from an Apple display, but a number of display manufacturers have signed up to do the same with USB-C. I'd love to be able to have a single cable to connect power, ethernet, display, keyboard and mouse to my laptop. About the only place where I connect more than one cable is at my own desk. I generally have to carry a dongle for projectors, because in most places they're VGA only (if you give a talk at Apple, they have an impressive array of adaptors connected to their projectors, for every display connector that Apple has ever sold. I really wished I'd lugged a G4 PowerMac tower with me to be able to use the ADC adaptor).

That's no different from closed source development. Do you think Microsoft or Apple care about the opinions of people who aren't going to buy their products? There is basically only one way to contribute to a closed source project: pay money to the developers. That works for open source too, but you can also produce code, documentation (please!), artwork, detailed and reproducible bug reports (please!). People who contribute in any of these ways are valuable to the project and their opinions should be considered. People who don't contribute anything are only valuable in the sense that they may eventually become contributors.

The difference between open source and closed development is that open source projects allow non-contributors to use the product for free, whereas most closed projects will use legal means to prevent this.

For HCI, this usually isn't difficult. There is a huge body of research that can be cited. Some of it changes. For example, we used to think that cancel buttons should go on the left and okay buttons on the right in left-to-right reading order countries, because the reading order influences how people perceive the direction that forward and backward buttons go. More recent research has shown that this doesn't depend on reading order: people perceive left as back and right as forward whatever their reading order. GNOME and Windows are particularly bizarre in this respect, putting okay on the left, cancel on the right, but having forward on the right and back on the left - they're not even consistent within the same dialog box!

As a FreeBSD developer, I have a lot of respect for the work Poettering. Every time he releases a new piece of software, we gain a load more users and developers. I can't wait for his next project.

Clearly you have never visited GitHub.

You might want to look at the std::array template in C++11. It lets you do exactly this, only with compile-time checking of the arguments that the lengths match. Actually, you can skip the A, B, C parameters because your function template would be able to extract them from the .size() of the matrix parameters.

This is a feature, not a bug. If uninitialised variables were default-initialised to some value, then this code would still be buggy, but now your compiler / static analyser would not be able to tell you that it contained a bug.

If you don't want that behaviour, then initialise the value at the declaration. Coding styles typically discourage initialising with a sentinel though, because you lose the ability for your compiler to check that you've assigned the correct value on every code path.

This is true. On Linux, the only compilers that I know of that use the same C++ ABI as gcc are clang, ICC, XLC, ARMcc, Open64 and Path64. If you're using something else then you may encounter problems.