Peter Dimov

What's the actual `Foo`?

No, `p` doesn't need to be aligned. Unless otherwise specified, a pointer to `unsigned char` doesn't have alignment requirements (because the alignment of `unsigned char` is always 1.) I'll make...

Ah, that's https://github.com/boostorg/context/pull/155. Well it breaks one of Fiber or Coroutine, then.

ASIO depends on Coroutine, so whether it's "deprecated" or not is of no relevance. It must work.

That's not my department. I'm just seeing that Fiber fails to build on Cygwin, and am telling you that it does, and why.

New Fiber failure after the latest changes: ``` [ 21%] Building CXX object libs/fiber/CMakeFiles/boost_fiber.dir/src/algo/algorithm.cpp.o In file included from /home/travis/build/boostorg/boost-root/libs/fiber/src/algo/algorithm.cpp:9:0: /home/travis/build/boostorg/boost-root/libs/fiber/include/boost/fiber/context.hpp:26:42: fatal error: boost/context/detail/apply.hpp: No such file or directory compilation terminated....

> fixed Indeed, thanks. > ASIO should switch to boost.context instead of using deprecated boost.coroutine. I've looked at your PR https://github.com/boostorg/asio/pull/55 that does exactly that, but it seems to me...

Throwing your own `bad_alloc` is ridiculous. `std::bad_alloc` is very often caught specifically by name, and is not even defined in ``.

These throw functions are much better placed in a compiled library (which would avoid the stdexcept include as a side effect), but this would probably be an unacceptable regression for...

See for instance the difference between the generated code in https://godbolt.org/z/4ThETbnM8 and https://godbolt.org/z/sE8P5s13h. When `BOOST_CONTAINER_USE_STD_EXCEPTIONS` is defined, according to CE's timings, the code actually compiles more quickly - 305 ms...