capnp_generate_cpp() checks that input schema files exist at configure-time, and reports a fatal error if they don't exist. However, the check prepended the value of CAPNPC_SRC_PREFIX to input file paths, which is the wrong thing to do: the input file paths should be checked as-is if they are absolute paths, and checked relative to the current source directory, NOT the value of capnp's src-prefix flag, if they are relative paths, in order to match the capnp tool's behavior.

It turns out that it's easiest to just unconditionally convert the input file paths to absolute paths, then check the absolute path. The reason is that we can't even pass relative paths to the capnp command, anyway: capnp interprets relative path inputs relative to its working directory, which defaults to the build dir. For consistency with other CMake commands (add_library, add_executable, etc.), it makes most sense if relative file path inputs to capnp_generate_cpp() are interpreted relative to the current source directory. This means that relative path inputs need to be converted to absolute paths before being fed to capnp, which was done right after the faulty existence check.

This commit fixes the issue by modifying the existence check to check the path only after it's been converted to absolute form.

Closes #586.

Extend UnixEventPort with the ability to listen for subprocess exit.

(Later, we should add a nice subprocess API around this, maybe even one that can work on Windows too...)

Improve stack trace accuracy by subtracting 1 from each return address.

This should prevent stack traces from spurriously pointing at the line after the one where the call actually happened.

This leaks due to the old compiler bug:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33799
http://llvm.org/bugs/show_bug.cgi?id=12286

Previously we only closed immediately after a response finished.

Add TaskSet.onEmpty() to wait until all tasks have completed

Fixes #286.

I intended to do this, like, 4 years ago.

Revert "avoid passing invalid reference to Array<T> constructor"

Memcpy null src

Revert "Don't read past the end of the base64 decode out buffer."

Don't read past the end of the base64 decode out buffer.

Don't bind Header reference to a null pointer.

Variable-length arrays must have positive bound.

It is invalid to pass null as a pointer argument to memcpy/memcmp/memset, even if the count argument is zero:

> Where an argument declared as size_t n specifies the length of the array for a
function, n can have the value zero on a call to that function. Unless explicitly stated
otherwise in the description of a particular function in this subclause, pointer arguments
on such a call shall still have valid values [...]

Detected by -fsanitize=nonnull-attribute.

If we finish decoding in step_a state, there is no current output character, so reading *plainchar will either be an uninitialized read or (if the output buffer is minimally sized) a past-the-end read.

Detected by -fsanitize=address.

If recvData is empty, recvData.begin() is a null pointer and binding a Header reference to it is invalid.

Detected by -fsanitize=null.

Detected by -fsanitize=vla-bound

Generate ASLR-friendly numeric stack traces.

Add Own<T>::attach() which is much like Promise<T>::attach()

We don't support any extensions currently, but it's important that when acting as a proxy, we don't pass through this header, since it could result in the client and server negotiating extensions that the proxy doesn't understand and will botch.