Notices by Alexander Monakov (amonakov@mastodon.gamedev.place)

@ignaloidas @wolf480pl Language committees never concerned themselves with sockets, and we didn't end up with such explosion. We also do not have a multitude of libraries for a simple TLS-wrapped socket abstraction.

I still think that in C and C++ (additions to) standard libraries should be just for things that cannot be properly implemented without coordination with the compiler.

Likewise, "breadth" of the stdlib is not a strength of the language, nor where the main weak points are.

@wolf480pl I'd say they can be a part of a normal library, not something the language committee needs to be concerned with. And how easy it's to write and maintain _is_ a strength of the language.

@dalias I do not see how this relates to [non-]transitivity. What would be the hypothetical third struct that would be compatible with exactly one of pthread_mutex_t and mtx_t?

Generally speaking, do you have an example that illustrates non-transitivity of compatibility relation that does not involve unprototyped function type?

Let's say a C program comprises two translation units, each of which declares the following type:

struct list {
struct list *next;
void *payload;
};

For deciding whether they are compatible, the following clause of the language standard applies:

"there shall be a one-to-one correspondence between their members such that each pair of corresponding members are declared with compatible types;"

Apparently the implementation may claim that such self-referential types are not compatible! 🙃

Now suppose the program contains not two, but three such translation units. Let's say the instances of the struct type are A, B, and C. Can the implementation choose to A be compatible with B, B with C, but C incompatible with A?

@wolf480pl I'm probably missing what you're hinting at, but to me the most "interesting" (but not really) question is what choice an implementation aiming to detect many instances of UB, such as TIS-interpreter, would make.

@icedquinn @wolf480pl I don't think it's academic, checking decoded data is how you catch unintended bugs; floating-point decoding can be reproducible if you know what you're doing

@wolf480pl @icedquinn I can confirm that, both the ffmpeg decoder and reference library use floats; the reference library can be configured to avoid floating-point arithmetic, but that's not the default.

@wolf480pl @icedquinn but even without floating-point, you can get different decodes if your other version of ffmpeg doesn't know about the clever backward-compatible upgrade:

@wolf480pl @icedquinn hm, but on second thought, that is specifically for streaming over networks, and wouldn't be used in conventional encoding?

@wolf480pl @icedquinn possibly? from the sound of it, the encoder would have to opt-in for that, and I barely know what I'm talking about here.

@wolf480pl @icedquinn again, on second^2 thought, in case of Youtube that wouldn't make sense, because they are streaming over TLS, which is not lossy. It is for use cases where the protocols are specifically built with dropouts in mind, like videoconferencing.

I am once again baffled that apparently nobody is caching random blobs like source tarballs, because only 'busybox wget' can issue 'GET https://...' to a proxy (wget and curl issue CONNECT for https resources), and Squid doesn't interoperate with 'busybox wget', as it sends TCP FIN early — I had to use nginx with 'proxy_ignore_client_abort on'.

But now we finally have a setup where we can have a caching http(s) proxy in front of containers that rebuild a distro.

In 'less', you can interactively add command-line arguments without leaving the pager by pressing '-': you can press '-S' to flip wrapping/chopping of long lines, and '-j11' to spawn 10 wor^W^W^W see extra ten lines of context above the match when searching!

@regehr how do you rate this verbatim quote from the paper:

GWP-ASan is neither GWP nor ASan.

@lanodan @wolf480pl good to know, thanks!

My kingdom for a distro with smooth cross-installation (preparing a rootfs for a foreign architecture without qemu).

@shafik and yet sometimes a quad-long is okay with g++ if you twist its arm in exactly the right way:

At work, I'm helping recover a server that fails to boot, and as I learn more about common UEFI implementations, I keep wondering "what were they thinking??"

- Modifiable state (NVRAM) on the same flash chip as essential boot software;

- No logs whatsoever.