fixed all tests on x86 GNU/voidlinux

[Plisp]

All tests now run on my system and hopefully everywhere else too. Summary of changes:

• according to the libuv docs the UV_FS_EVENT_RECURSIVE flag is not supported on linux. However to my pleasant surprise, providing 0 worked on directories anyways.
• fixed improperly indented doc line from my previous pull req
• use alexandria:ensure-list in util/helpers.lisp
• google's public dns has changed
• fixed fs-monitor test - after fixing the flag issue, the problem was that the fs handle was keeping the loop alive, causing a timeout. Also removed possibly spurious as:delay surrounding the checking code.

This closes https://github.com/orthecreedence/cl-async/issues/139 and https://github.com/orthecreedence/cl-async/issues/137. Also note that https://github.com/orthecreedence/cl-async/issues/65 is possibly already solved.

[Plisp]

Removed :export symbols for deprecated fd api. https://github.com/orthecreedence/cl-async/issues/63 is irrelevant with pollers (and tests) implemented and may be closed.

[Plisp]

Hmm...in ssl/tcp.lisp, the note on attach-ssl-to-socket says Note that this function *really* should be called before the passed socket is connected, and certainly before any data is exchanged over it. which is alarming since on L402, before the connection happens, ssl-run-state is called, which could (maybe) call write-socket-data with a non-nil :end kwarg, which would pass that arg to streamish-write, dispatched on ssl-socket. However write-ssl is NIL, control flows to call-next-method on L99, calling streamish-write dispatching on socket. Now since the socket is not connected, and the user has switched *buffer-writes*, control passes to call-next-method on L165, calling write-to-uvstream, which ignores the :end argument. Doh!

[Plisp]

Oh nvm that call's in the wrap-connect-cb which is executed strictly after setting socket-connected to T. You were right ;) Oh well at least the style-warnings are gone.

[orthecreedence]

Thank you!