iphydf

@fuhsnn https://github.com/TokTok/c-toxcore/pull/2624 it works perfectly. I'm adding it to our CI to make sure it keeps working.

No, all other C++ standard library headers can be found. Also, `-x c++` was already passed by the toolchain. I can see: ``` /nix/store/y02xngjcbhfgliijh2wnpksc3dqhva56-clang-wrapper-16.0.1/bin/cc -U_FORTIFY_SOURCE -U_FORTIFY_SOURCE -fstack-protector -Wall -Wthread-safety -Wself-assign...

The problem is that Go stdlib uses that (when compiling with msan) but doesn't declare it, because it's assumed to be part of the toolchain. Same with absl and probably...

I've moved away from `__dealloc__` entirely, and use `__del__` only. Works for me, but I still wanted to report it.

This is the ubsan stack trace, maybe that helps: ``` ../src/util/migrate.cpp:449:32: runtime error: member call on null pointer of type 'namespacet' SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior ../src/util/migrate.cpp:449:32 in UndefinedBehaviorSanitizer:DEADLYSIGNAL ==8==ERROR: UndefinedBehaviorSanitizer: SEGV...

Instead of casting, can we add flags to the function signature to express these properties? If it's cast, I need to make a macro to wrap my allocator function, which...

Hmm true, if it's returned through an out-parameter, it gets trickier. Not sure how to design that yet.

I'll try with the manual static_set for now. I'll see how far I get. Btw I'm trying to make https://github.com/TokTok/c-toxcore work with cake. It's a 65kloc C library, and my...

Ok, I've reverted the `CMP_NULL` change for now, keeping the prefix-incr change and moving a local variable to a deeper scope (still C89 compatible).

See http://clang.llvm.org/docs/BlockLanguageSpec.html.